Remote Surveying of Underground Cavities.pdf

Remote Surveying of Underground Cavities

Excavated by Jet Boring

by

Carolyn Ann Ingram

A thesis submitted to the

Robert M. Buchan Department of Mining

in conformity with the requirements for

the degree of Master of Applied Science

Queens University

Kingston, Ontario, Canada

September 2014

Copyright Carolyn Ann Ingram, 2014

Abstract

Cigar Lake is a high-grade uranium deposit, located in northern Saskatchewan, Canada.

In order to extract the uranium ore remotely, thus ensuring minimal radiation dose to

workers and also to access the ore from stable ground, the Jet Boring System (JBS)

was developed by Cameco Corporation. This system uses a high-powered water jet

to remotely mine out cavities. Survey data is required to determine the final shape,

volume, and location of the cavity for mine planning and development.

This thesis provides an overview of the challenges involved in remotely surveying

a JBS-mined cavity. In particular, it studies range finding sensors that are relevant to

mining applications and their attributes. As an alternative to sensors used for remote

cavity surveying, it evaluates the potentially advantageous features of a time-of-flight

(ToF) camera.

Data was collected from inside a test cavity in a variety of experimental envi-

ronments meant to simulate conditions in a real Cigar Lake cavity. Field data was

collected from the core shack at Cigar Lake and from an open stope at Rabbit Lake.

Advanced data analysis techniques such as registration and segmentation are also

explored for application in cavity surveying. The data from the ToF camera was

evaluated with respect to the survey systems slated for use at Cigar Lake and the

advantages for its use in the post cavity survey are shown.

i

Acknowledgments

Firstly, I would like to extend a multitude of thanks to my research supervisor, Dr.

Joshua Marshall. I feel blessed to have had a supervisor, and friend, who was positive

and encouraging, offering countless valuable ideas and providing continual guidance

throughout the journey.

I wish to convey my gratefulness to Cameco for the support I have received, not

only in funding for my studies, but also towards my professional development goals.

In particular, I would like to thank Martin Wacker for supporting my graduate study

ambitions and interest with regards to the jet boring system. Additionally, I owe a

big thank you to the JBS engineering crew, Devon Loehr, Dustin Repski, and Sean

Borycki, along with the Mining and Geology groups at Cigar Lake who provided

invaluable information and assistance over the past three years.

I would like to thank the Saskatchewan Research Council, specifically Kim Young,

Steve Kosteniuk, Nathan Peter, Damian Rohraff, and Ken Babich, for generously

providing me with an office and the space required to build the test cavity. They

showed an ever readiness to lend an extra hand, brain, or equipment and I was

extremely appreciative for their great patience as I made a mess with stucco, paint,

and water.

This research was funded in part through a Natural Sciences and Engineering

ii

Research Council of Canada (NSERC) IPS 1 scholarship.

Finally, I must express a deep sense of gratitude towards those people who love me

and keep me sane; my mom, Eleanor Ingram, my sisters, Melinda Zerr and Jennifer

Welsh, and all my faithful friends. I am so thankful that they found ways to spend

time with me (even if it meant slaving in the test cavity) while I struggled to find a

balance between work, study, and a seemingly elusive life.

iii

Contents

Abstract i

Acknowledgments ii

Contents iv

List of Tables vi

List of Figures vii

Nomenclature x

Chapter 1: Introduction 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Scope of Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Format of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Chapter 2: Cigar Lake Cavity Scanning 72.1 Survey System History . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1.1 Jet Boring System . . . . . . . . . . . . . . . . . . . . . . . . 72.1.2 Cavity Survey System . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Chapter 3: Theory and Background 163.1 Underground Range Measurement . . . . . . . . . . . . . . . . . . . . 163.2 Sensor Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.2.1 Ultrasonic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.2.2 Laser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.2.3 Radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.3 Cigar Lake Sensor Selection . . . . . . . . . . . . . . . . . . . . . . . 243.3.1 Historical Options Analysis . . . . . . . . . . . . . . . . . . . 24

iv

3.3.2 Time-of-Flight Camera . . . . . . . . . . . . . . . . . . . . . . 273.3.3 Device Comparison . . . . . . . . . . . . . . . . . . . . . . . . 30

3.4 Point Cloud Processing . . . . . . . . . . . . . . . . . . . . . . . . . . 323.4.1 Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.4.2 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Chapter 4: Experimental Studies 364.1 Test Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364.2 Test Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.3 Test Environments and Results . . . . . . . . . . . . . . . . . . . . . 40

4.3.1 Baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.3.2 Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3.3 Freeze pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464.3.4 Fog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.4 Field Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564.5 Point Cloud Processing Results . . . . . . . . . . . . . . . . . . . . . 57

4.5.1 Registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.5.2 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

Chapter 5: Summary and Conclusions 655.1 Summary and Contributions . . . . . . . . . . . . . . . . . . . . . . . 655.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Bibliography 74

Appendix A: MATLAB Code 84A.1 Main Script for SwissRanger and Senix Scan . . . . . . . . . . . . . 84A.2 Function to Obtain Data from SwissRanger . . . . . . . . . . . . . . 86A.3 Function to Obtain Data from Senix Sensor . . . . . . . . . . . . . 88A.4 Function to Correctly Format Celestron Command . . . . . . . . . . 88A.5 Function to Save SwissRanger Data in .pcd Format . . . . . . . . . . 89

Appendix B: Equipment Specification Sheets 92B.1 Senix ToughSonic TSPC-30S1 . . . . . . . . . . . . . . . . . . . . 92B.2 C-ALS MK3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95B.3 MESA SwissRanger SR4000 and SR4030 . . . . . . . . . . . . . . . . 100

v

List of Tables

3.1 Comparison of survey device specifications. . . . . . . . . . . . . . . . 33

vi

List of Figures

2.1 JBS mining schematic for Cigar Lake (image courtesy of Cameco Cor-

poration). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 Original prototype laser scanning tool (image courtesy of Cameco Cor-

poration). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.3 Computer rendering of mined cavities (image courtesy of Cameco Cor-

poration). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.4 Deployment tool testing at McArthur River mine site, June 2012 (im-

age courtesy of Cameco Corporation). . . . . . . . . . . . . . . . . . . 14

2.5 Vertical rod pusher deployment depth versus tool weight. Prototype

testing conducted in backfill pipe with 10.16 cm I.D. on a 70 incline

at McArthur River mine site. . . . . . . . . . . . . . . . . . . . . . . 15

3.1 Electromagnetic spectrum [1]. . . . . . . . . . . . . . . . . . . . . . . 17

3.2 Target reflective characteristics [6]. . . . . . . . . . . . . . . . . . . . 22

3.3 Time of Flight sampling of returned modulated signal [5]. . . . . . . . 31

3.4 Devices compared for Cigar Lake cavity surveying. From left to right:

MESA SwissRanger (ToF Camera), MDL C-ALS (Laser Scanning

Tool), Senix ToughSonic (Ultrasonic Sensor). . . . . . . . . . . . 31

4.1 Design and construction of test cavity. . . . . . . . . . . . . . . . . . 37

vii

4.2 C-ALS test apparatus in test cavity. . . . . . . . . . . . . . . . . . 38

4.3 SwissRanger camera Cartesian coordinate system, (x, y, z) [5]. . . . . 40

4.4 C-ALS baseline scan (vertical) demonstrating data acquisition time. 42

4.5 C-ALS vertical scan 3D plots shown with increasing acquisition in-

tervals (colour scaled by signal strength with blue for low and red for

high). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.6 Left: Basic image acquisition from SwissRanger using MATLAB

(amplitude (grayscale) image, range image, and confidence map). Right:

Point cloud data plotted using the PCL viewer. . . . . . . . . . . . . 45

4.7 SwissRanger Point Cloud data with water (blue) and without water

(green) on test cavity surface. . . . . . . . . . . . . . . . . . . . . . . 47

4.8 SwissRanger Data with Water Droplet on Lens. . . . . . . . . . . . . 48

4.9 C-ALS scan of freeze pipes. Scanning interval (Left to Right): 6,

3, and 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.10 SwissRanger data with freeze pipes in test cavity. . . . . . . . . . . . 50

4.11 Inside of test cavity with light, medium, and dense fog conditions. . . 51

4.12 C-ALS data in dense fog (left) and baseline (right). Target distance

shown on horizontal axis in meters. . . . . . . . . . . . . . . . . . . . 52

4.13 Point cloud data from SwissRanger in light, medium, and dense fog

conditions inside the test cavity. . . . . . . . . . . . . . . . . . . . . . 54

4.14 Amplitude images from SwissRanger in fog (Auto-scaled in MATLAB). 55

4.15 Side view of point cloud data from SwissRanger acquired from posi-

tions 2 m apart in no fog (top) and medium fog (bottom) conditions.

Integration time is increasing from left to right. . . . . . . . . . . . . 56

viii

4.16 Field data acquisition with SwissRanger at Cigar Lake and Rabbit Lake. 58

4.17 SwissRanger images of Cigar Lake core sample. . . . . . . . . . . . . 59

4.18 SwissRanger images acquired at an open stope at Rabbit Lake Mine. 60

4.19 Registration applied to SwissRanger images in test cavity without po-

sition information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

4.20 Segmentation applied to intensity point cloud. . . . . . . . . . . . . . 64

ix

Nomenclature

C-ALS Cavity Auto-Scanning Laser MK3

Interim Cavity Survey Survey performed during cavity excavation between peri-

ods of jetting

JBS Jet Boring System

LIDAR Light Detection And Ranging

PCL Point Cloud Library

Post Cavity Survey Survey performed after cavity is excavated

RADAR Radio Detection And Ranging

Senix Ultrasonic Sensor Senix TSPC-30S1 ToughSonic Distance Sensor

SONAR Sound Navigation And Ranging

SRC Saskatchewan Research Council

SwissRanger MESA Imaging Swiss Ranger ToF Camera, SR4030

ToF Time of Flight

x

Vertical Rod Pusher System developed to deploy post cavity survey device at

Cigar Lake

xi

1Chapter 1

Introduction

In mining, surveying is a common task where data collection takes place using com-

mercially available technology. For unique circumstances, such as the one at Camecos

Cigar Lake mine site, the available technology may have to be altered to suit a

particular need. Because of Cigar Lakes unique deposit, a correspondingly unique

method of mining was developed to extract ore and, as might be expected, requires

a unique method of surveying to observe the mined areas. This dissertation provides

an overview of the basic technologies that have been used for range finding in mining

and focusses on the potential for use of a time-of-flight camera (see Section 3.3.2) in

application at the Cigar Lake mine site, with comparison to the currently proposed

systems, for surveying cavities.

1.1 Motivation

Cigar Lake is the worlds second largest known high grade uranium deposit and is

located in northern Saskatchewan, 40 km inside the margin of the eastern part of the

Athabasca Basin. The Cigar Lake deposit is approximately 1 950 m long, 20 to 100 m

wide, with an average thickness of about 5.4 m. It occurs at depths ranging between

1.1. MOTIVATION 2

410 and 450 m below the surface. The body of high grade mineralization located at

the unconformity contains the bulk of the total uranium metal in the deposit and

currently represents the only economically viable style of mineralization, considering

the selected mining method and ground conditions [10].

The jet boring system (JBS), a remote mining method, was developed to extract

the Cigar Lake deposit. The unique challenges involved in accessing the deposit in-

clude ground instability, radiation protection, control of groundwater and a relatively

thin, flat lying mineralization. The JBS was selected after many years of exploration

and test mining activities, following the discovery of the deposit in 1981. The JBS

mining method consists of remotely excavating cavities out of frozen ore with a high

pressure water jet, producing approximately 230 t of ore per cavity. An average cav-

ity is expected to be approximately cylindrical in shape, with diameter of 4.5 m and

a depth of 6.0 m. Although Cameco has successfully demonstrated the JBS mining

method in trials, at the time of writing, this method had not been proven at full

production.

Due to the unproven nature of the mining method, it is expected that many tech-

nical issues may arise as ramp-up to full production progresses. For the purpose

mining optimization and understanding how parameters set during the process of

jetting affect the outcome of the excavated cavity, it is necessary to have a reliable

method for surveying and interpreting data acquired from inside the cavity. In pro-

totype testing completed with a cavity survey system in 2000, which is described in

Chapter 2, the systems used showed to be unreliable due to issues with both the

sensor and telemetry method. For this reason, it was suggested that further research

into options for surveying the cavities could prove to be of value, as production began

1.2. PROBLEM FORMULATION 3

at Cigar Lake.

1.2 Problem Formulation

The testing that took place with the JBS in 2000 was the single opportunity for

testing all components of the system in the field. Many valuable lessons were learned,

and areas for improvement identified. A critical component for identifying the success

of the JBS, in the cavity excavation process, is the survey system. Throughout the

eight cavities that were excavated, four in waste rock and four in ore, attempts to

use the survey system, as designed and with modifications, were made. The primary

issues identified with the survey system were:

1. Communication or telemetry

2. Mechanical robustness

3. Unreliable range data

At the time of writing, the Saskatchewan Research Council (SRC) had already

begun to address the issues of telemetry and mechanical robustness for the interim

survey system. They had also tested and chosen an ultrasonic sensor for range find-

ing during the jetting process. However, no focus had been made on assessing the

technology available for the post-cavity survey, which was intended to collect detailed

data after the cavity was complete. Although the extreme environment created by

jetting would not have as large of impact for the post-cavity survey as it would for

the interim, it was still possible for fog, falling debris, water, frost, freezepipes to be

present in the cavity. Additionally, there were size limitations imposed by the de-

ployment method which would restrict potential options for surveying the remotely

1.3. SCOPE OF WORK 4

excavated cavities.

It was proposed that a review of technologies for range finding, the sensor charac-

teristics and applicability for mining applications be conducted. A further assessment

into which technologies on the market would have capability for use as a survey tool

in the Cigar Lake cavities could then occur and would be based on cost, physical

limitations and ability to withstand the cavity environment. The scope of the work

completed for this dissertation is described in the following section.

1.3 Scope of Work

Given the unique challenges involved with accessing the remote cavities at Cigar

Lake, the limited testing that has taken place, and the extreme environment that

surveying is to occur within, the scope of this dissertation was created to encompass

the following objectives:

1. A review of documentation relating to the cavity survey systems tested at Cigar

Lake, with the purpose of identifying the successes and shortfalls of the system,

along with areas for improvement.

2. A review of basic range sensors and the characteristics that can affect their

performance in adverse conditions relating to a mining environment and, more

specifically, for the Cigar Lake application.

3. Evaluate the ToF camera as an alternate technology in comparison to the ul-

trasonic and laser-based devices for use at Cigar Lake.

4. The design and construction of a test area and apparatus for the three devices

1.4. FORMAT OF THESIS 5

(ultrasonic, laser, and ToF camera) along with the implementation of exper-

iments designed with intention to test and compare the device performance

within isolated environmental conditions expected inside the Cigar Lake cavi-

ties.

5. Collect data, as possible, in the field with a ToF camera and evaluate it for

application in cavity surveying.

As a result of the experiments and data collection with the ToF camera, it was

possible to consider advanced data analysis techniques, which include registration

and segmentation (see Section 3.4). It is proposed that analysis could provide addi-

tional information that is not possible with the laser and ultrasonic survey systems.

With further research, and field data collected inside a Cigar Lake cavity, it would

be possible to confirm their utility in the field, but falls outside the scope of this

dissertation.

1.4 Format of Thesis

The next chapter provides an introduction to the Jet Boring System (JBS) that is

used at Cigar Lake and a summary of the testing which had previously taken place

with the cavity survey system. It outlines the lessons learned, the conditions to expect

inside the cavity and the limitations which are imposed by the the access and methods

possible to place the survey device inside a cavity and transmit the acquired data.

Chapter 3 reviews the basic properties of ultrasonic, laser, and radar signals.

It describes how these signals may be affected in a mining environment and the

differences between them that could influence the decision making process. A MESA

1.4. FORMAT OF THESIS 6

SwissRanger ToF camera is also presented as an alternative to the ultrasonic and laser

systems that are slated for use in the cavity surveying application at Cigar Lake.

In Chapter 4, the various experimental test environments and apparatus for ob-

serving the differences between the Senix ultrasonic sensor, MDL Cavity Auto-

Scanning Laser (C-ALS) system, and the MESA SwissRanger ToF Camera are

described and include the respective settings used for each device. The test environ-

ments were with water, freezepipes, and fog present in the test cavity. Data was also

collected on site at Cigar Lake mine and Rabbit Lake mine in effort to show how data

collected in the field may appear. An analysis of the data from each experiment and

location is presented.

Finally, a summary of the results of this dissertation and the associated conclusions

are given in Chapter 5. Based on the results obtained, suggestions for future work are

presented and could be used in further development of a 3D cavity surveying system

with a ToF camera.

7Chapter 2

Cigar Lake Cavity Scanning

2.1 Survey System History

This section provides a brief background and history of the jet boring system (JBS)

and the development of a scanning survey system for the excavated cavities at Cigar

Lake.

2.1.1 Jet Boring System

At Cigar Lake, the Jet Boring System (JBS) was developed to access a high-grade

uranium orebody that is situated in water-saturated sandstone. Before mining begins,

the orebody and surrounding rock is frozen to strengthen it and also to prevent inflows.

A pilot hole is then drilled up through the ore body and cased, providing a path for

the jet string and nozzle. The jetting begins at the top of the ore body and progresses

downwards in periods, as the jet rotates and traverses about its axis, until the lower

limit of the ore body is reached. As the cavity is jetted, the ore slurry falls through

the annulus of the pilot hole casing and jet pipe and into a slurry storage tank before

being pumped to the run-of-mine (ROM) area. From here, further processing of the

2.1. SURVEY SYSTEM HISTORY 8

slurry occurs underground before being finally pumped to surface [10].

The purpose for the cavity survey, which will take place between periods of jetting,

is to provide feedback to the operator, who is located beneath the orebody (see Figure

2.1), indicating the dimensions of the cavity. While mining, it is important to achieve

maximum recovery while preventing ground instability that may be caused by too

large of a cavity. Following the mine plan, the operator will need to know if it is

necessary to focus the jet on a particular area within the cavity, change the jetting

parameters, or to cease mining. Additionally, the final location of the cavity and its

dimensions are required in order to update resource estimates and production values.

The volume of required backfill can be also be determined if the cavity shape is

known. Knowledge of the JBSs performance and the ability to optimize procedures

to ensure efficient maximum recovery are necessary for a successful mining program.

The challenge is to find an appropriate technology that can acquire reliable data

within the extreme cavity environment.

Figure 2.1: JBS mining schematic for Cigar Lake (image courtesy of Cameco Corpo-ration).


2.1.2 Cavity Survey System

At Cigar Lake, the original goal was to precision mine each ore cavity. The first

prototype (see Figure 2.2) contained several features, including a laser rangefinder, an

IR target detector, IR LEDs for camera illumination, and sapphire lenses for the laser,

video, and target detector. As a result of the field testing described below, better

knowledge of the operating conditions was obtained, and changes to the original

specifications and prototype were made. Field testing of prototype cavity survey

systems took place in 2000 from April through November, in which eight total cavities

were created, four in waste rock and four within the ore body [63].

Figure 2.2: Original prototype laser scanning tool (image courtesy of Cameco Corpo-ration).

The first set of preliminary tests were conducted within a vertical culvert in an

underground raise, with simulated rock conditions similar to those expected during a

typical mining situation. Through the process, it was found that the level of control

to cavity shape with jetting was less than anticipated, and thus, the inclusion of


survey components such as the target detector and video camera, to view the cavity

shape during mining, would not provide value. With the system simplification, it was

possible to move the laser range finder higher above the jetting nozzle and further

from the debris that may be launched as a result of jetting. The next phase of testing

took place in frozen waste rock conditions below the orebody. With water being

sprayed in a frozen environment, a resultant dense fog ensued and caused the laser

rangefinder cavity survey system to be rendered inadequate. Some effort was made

in an attempt to eliminate or reduce the fog, by means of a Transvac vacuum and

through use of compressed air, but both were unsuccessful.

An ultrasonic sensor was subsequently chosen as the replacement for the laser

rangefinder due to its off-the-shelf availability, low cost, and, most importantly, the

ability of the signal to produce an indication of range within the fog. During testing

in the ore body, it was discovered that significant cavity wall erosion occurred above

the jet as mining progressed downward and, as a result, it was determined that a final

survey of the cavity would be required prior to backfilling. Preferably, this survey

would be completed as quickly as possible after jetting was complete, before any

additional sloughing occurred from the potentially unstable walls of the excavated

cavity. Since the ultrasonic sensor did not provide a high enough level of precision

and resolution for mine planning purposes, and the data processing system had not

been developed to the point where estimates of the cavity volume could be calculated,

the decision was made to again employ the use of a laser rangefinder. The post-cavity

survey would be in addition to the ultrasonic survey, which would still be completed

between periods of jetting to provide feedback regarding the growth of the cavity

radius to the operator. Through the process of field testing, it was found that the


Cigar Lake cavity shape could be approximated by a cylinder of height ranging from 3

m to 15 m and a diameter ranging from 3 m to 5 m, with the cavity surface comprising

of dark, wet, frozen uranium ore.

Figure 2.3: Computer rendering of mined cavities (image courtesy of Cameco Corpo-ration).

The 2000 JBS Test Report [63] identifies the inconsistent performance of the

survey system as an area requiring improvement, one that had caused a significant

impact to cycle times. The delays were ultimately caused by a number of issues which

negatively impacted the survey system performance and are summarized in the list

below.

1. Communication or telemetry: Must be able to consistently access data from

survey system


2. Mechanical robustness: Must withstand impact from debris, pressurization

of cavity (if plugging occurs), and vibration during jetting. Survey device must

be operate after submersion in water.

3. Unreliable range data: Potential causes of inaccurate data are suspect from

water or dirt on lens and the cavity fog. Method to diagnose issues is beneficial.

Impact was noted on inaccurate calculation of backfill volumes.

In 2005, a review of current technology was conducted by the Saskatchewan Re-

search Council (SRC), in which a Cavity Auto-Scanning Laser (C-ALS) system1

was identified at the most viable technology for the post-cavity survey system [37].

It was small enough to fit inside the backfill pipe (12.7 cm inner diameter) and was

equipped with a suitable deployment method and data processing software (see Ap-

pendix B.2). For the interim cavity survey system, in 2006, several ultrasonic range

sensors were tested in lab, where the Senix ToughSonic TSPC-30S1 (see Ap-

pendix B.1), demonstrated the best performance for the Cigar Lake application [43].

As part of a continuing contract with Cameco, SRC has researched and developed

an improved housing and electronics design for increased mechanical robustness of

the interim cavity survey system and has also developed two telemetry options for in-

creased reliability of data acquisition. These two telemetry options include: the Power

Line Modem (PLM) method which utilizes the isolated inner and outer pipes of the

jet string, and the Acoustic method which transmits signals acoustically through the

jet string.

1See http://www.renishaw.com/en/c-als-borehole-deployable-laser-scanner-for-concealed-cavity-and-void-scanning--25590, accessed on August 9, 2014.

2.2. DEPLOYMENT 13

2.2 Deployment

In addition to the limitations incurred by the environment in which the scan must

take place, the method of deployment also affects the size and weight of the tool which

would be used to survey the remote cavity. The laser survey tool which was used to

obtain the final cavity scans in the 2000 testing was found to provide the necessary

data but the manual method of deploying the tool up-hole was cumbersome, time

consuming, and arguably unsafe.

Initially in October 2010, and again with a rod of increased stiffness in June of

2012, a custom deployment method called the Vertical Rod Pusher was tested at

the McArthur River mine site [25] as an alternative to the manual method. The

custom device was designed for pushing a continuous length of specialized semi-rigid

fiberglass rod through conduits by using a motor with two opposing drive tracks that

clamp down on the rod.

In order to deploy the laser survey tool into the cavity, the jet string must be

removed from the hole, and the backfill pipe installed to provide a necessary conduit

for the rod and tool. An operator is also required to retrieve and position the JBS

cassette which houses the deployment system components. Based on cycle times from

the initial commissioning of the JBS in 2013, the aforementioned process would take

approximately 12 hours. The operator is required to make the necessary connections

between the rod, survey tool, and conduit before simply actuating a valve to deploy

the tool. The backfill pipe at Cigar Lake, with a relatively small inner diameter of

12.7 cm, defines the maximum size of any survey device that is deployed using this

tool.

Identical conditions to those expected at Cigar Lake could not be achieved in

2.2. DEPLOYMENT 14

Figure 2.4: Deployment tool testing at McArthur River mine site, June 2012 (imagecourtesy of Cameco Corporation).

the field at McArthur River mine site, but the prototype testing showed definite

promise for the Vertical Rod Pusher in deployment of lightweight survey tools. Upon

completion of data analysis, it was determined that, in order to achieve deployment

up to 60 m in a 10.16 cm I.D. hole at a 70 incline, the post cavity survey tool must

weigh less than 17.8 kg (see Figure 2.5). Since Cigar Lakes backfill pipes would have

a larger inner diameter of 12.7 cm, with more potential for snaking inside the pipe,

and could be oriented at inclines of 90 or less, there was potential that the maximum

deployment distance might diminish, if the weight of the tool was not also reduced.

Thus, this weight limitation was considered in evaluating potential post-cavity survey

devices.

2.2. DEPLOYMENT 15

Figure 2.5: Vertical rod pusher deployment depth versus tool weight. Prototype test-ing conducted in backfill pipe with 10.16 cm I.D. on a 70 incline atMcArthur River mine site.

16

Chapter 3

Theory and Background

3.1 Underground Range Measurement

Range measurement can be accomplished by several methods but, due to the access

restrictions for the Cigar Lake cavity application, this section focuses on sensors that

can be operated remotely. Some common acronyms referring to systems used to mea-

sure range are SONAR (Sound Navigation And Ranging), LIDAR (Light Detection

And Ranging), and RADAR (Radio Detection And Ranging). The basic principle

underlying the measurement of distance for all these devices is the time-of-flight of

the signal. The signal is sent using a transmitter and detected by a receiver. With

knowledge of the speed of the signal, and how long it took for the signal to travel to

the point of interest and back, you can determine the distance it travelled (i.e., the

target range). The basic equation describing this relationship is

t =2d

v, (3.1)

where t is the delay time from when the signal was transmitted to when it was

received, d is the distance to the target and, v is the speed of the signal.

3.1. UNDERGROUND RANGE MEASUREMENT 17

Four different types of sensors are examined in the following section. These are:

ultrasonic,

laser,

microwave radar, and

millimeter wave radar.

An ultrasonic sensor uses a sound wave as a signal and the laser, microwave radar,

and millimeter wave radar all use light, and thus, fall within the electromagnetic

spectrum, as shown in Figure 3.1.

Figure 3.1: Electromagnetic spectrum [1].

The physical differences between signals that are a part of the electromagnetic

spectrum lie only in their wavelength and frequency. It is the state of the technology

for devices which produce and interpret the signals, and how the signals behave when

3.2. SENSOR OVERVIEW 18

interacting with the environment, that differentiates how applicable they may be

within the context of mining and at Cigar Lake.

3.2 Sensor Overview

3.2.1 Ultrasonic

An ultrasound wave is a longitudinal, mechanical wave, where the accuracy in mea-

suring its speed, and from that, the distance it travelled, relies on knowing the speed

of sound in the medium. A sound wave that can be heard by humans lies in the

frequency range between 20 Hz and 20 000 Hz [46] and thus, an ultrasound wave

typically has a frequency greater than 20 000 Hz. The relationship between the fluid

(gas or liquid), which the signal will propagate through (i.e., the medium), and the

speed of the signal can be described by,

v =

B

0, (3.2)

where v is the velocity of sound (m/s), B is the bulk modulus (Pa) and, is the

density of the fluid (kg/m3) [46]. Inaccuracies in range measurement will arise if

the medium is not homogeneous or if compensation for changing conditions is not

accounted for. For example, at room temperature, a change of 10C will result in

approximately 1 % change in the speed of sound and thus the measured distance as

well [3]. Often, temperature compensation is a feature of ultrasonic range sensors,

but would only account for temperature changes at the sensor itself. At Cigar Lake,

it is expected that the temperature at the sensor would be much higher than the

frozen cavity wall, and so the temperature gradient within the air space between,


could cause inaccuracies.

Additionally, there are effects on the accuracy of the range measurements that

are caused by external influences. Since the medium is the carrier of the wave, a

bulk movement of the medium by an air current will displace the acoustic wave, and

furthermore, if the airflow becomes turbulent, it will cause a disruption in both the

transmitted pulse and the echo, resulting in severe attenuation. External vibrations

affecting the sensor may cause a shift in the carrier frequency that would result in

reduced sensitivity. Depending on where the sensor is mounted and what sort of

ventilation exists, these could become factors in the accuracy of the measurement

within an underground mine [6].

Another consideration regarding the effectiveness of the range measurement must

be the properties of the target. All materials will partially reflect, partially absorb,

and partially transmit the incident wave. The proportion of energy reflected is a

function of the ratio of the characteristic impedance of the solid target to the medium

through which it travelled. Hard or dense targets tend to reflect well, while soft targets

would tend to transmit or absorb.

The geometry of the area surrounding the target and the angle of incidence to the

target will also have an effect on the acquired measurements [6, 56]. It may be possible

for the echo to be reflected away from the transducer and/or be reflected at multiple

points before returning to the receiver. An ultrasonic signal also has a relatively wide

beam width (lower frequency = wider beam width) and, if it were to strike an areacomposed of several distances it would be difficult to resolve the correct component

of the returned signal [36]. Evidently, an awareness of the geological composition of

potential targets underground as well as their surface structure must be known in


order to have a well balanced knowledge of how the ultrasonic sensor will perform.

Ultrasonic sensors have been tested in underground mining applications in the

past [2, 56, 52] and have shown to be able to withstand extreme environments which

include vibration, dust, and fog. Their relatively inexpensive cost and availability also

make them an attractive range sensing option. However, the level of precision and

accuracy required by a particular application may encourage the pursuit of a different

technology, as was deemed necessary by Cigar Lake for their final cavity survey (see

Section 2.1.2).

3.2.2 Laser

A laser emits light in the visible range of the electromagnetic spectrum. It has the

shortest wavelength (approximately 380 to 760 nm) and the highest frequency (400 to

790 THz) of the signals that are discussed. The word laser is an acronym for, Light

Amplification by the Stimulated Emission of Radiation, which hints at the process by

which a laser signal is created. An electron can be excited to a higher energy level,

and when it returns to its stable, lower energy state, the energy is released in the

form of a photon. These photons are the constituents of a laser beam and, since the

photons are released at a particular energy, the resultant signal can be made highly

monochromatic, coherent, and directional [27].

In contrast to ultrasonic sensors, the conditions of the environment such as tem-

perature, pressure, air currents, will not have an effect on the propagation velocity

(i.e., the speed of light). The simple relationship between the signals propagation

speed, v, and the speed of light, c, is


v =c. (3.3)

The relative dielectric constant, , is the only property of the medium that would

have an effect on its speed. For the short range application that is considered in

this study, the effect of atmospheric attenuation, due to molecular interactions with

electromagnetic radiation would be small, and the value of near one for air.

It is possible to obtain high angular resolution and long range measurements with

a laser, however, the accuracy and range are highly dependent on visibility within the

medium and the targets properties. As an electromagnetic signal propagates through

the atmosphere, molecular interactions with the wave will absorb energy, and the

signal amplitude will decrease as the range increases. In clear air the attenuation is

minimal but for mining applications, where dust particulates or fog are often present,

the attenuation could be severe [51, 12, 6] and would have dependency on the par-

ticle size and visibility. If the laser signal interacts with the particulates in the air,

spurious readings may also occur as photons are returned prior to the beam reaching

its intended target.

Like the ultrasonic sensor, the properties of the target itself also affect the per-

formance of the laser sensor. With a highly reflective, diffuse scattering target, the

best performance will be observed (see Figure 3.2). Conversely, low reflective, diffuse

scattering surfaces may absorb the signal and will not be effective in measuring the

range. On a smooth, shiny or wet surface, specular reflection may occur and depend-

ing on the angle of incidence, the reflected beam may not be returned to the receiver.

For the case of retro reflection, one will get a measurement of high intensity but this

is not necessarily a positive result as it may saturate the receiver.


Figure 3.2: Target reflective characteristics [6].

With the maturity of laser technology, it has already found use within several un-

derground mine mapping applications [59, 40, 8, 33]. However, it is the environment,

with varying levels of dust and humidity, which is clearly acknowledged to limit the

performance of the device in use. A comparison of laser and radar ranging devices,

within adverse environmental conditions meant to simulate an underground mine,

was carried out in [51] and demonstrated the limitations that may be encountered.

Differences were found between the performance of the laser technologies as a result

of the signal wavelength and data processing technologies. It was suggested that none

of the sensors could alone be relied upon in the mining application for which they

examined.

3.2.3 Radar

Microwave Radar

Like the laser, the microwave radar signal is also composed of an EM wave. It has a

frequency which falls in the range between approximately 30 kHz and 30 GHz and a


wavelength of between 10 mm and 10 m. As alluded to in the previous section, mea-

surements will be affected by the concentration and size of particulates in the medium.

It should be noted that the absorption or attenuation effects only become severe as

the wavelength approaches the size of the suspended particle. The wavelength of

microwave radar is much larger than the diameter of typical dust or water vapour

and so, although a significant factor when using a laser, not generally a problem for

radar [11].

An additional consideration must be again the beam width. The lower the fre-

quency, the larger the beam width of the signal will be. Over a long range, the beam

will disperse and can become a significant issue, potentially reflecting off unintended

targets [7]. Over a short range, the dispersion effect is less severe and still offers an

improvement over the ultrasonic option with respect to resolution. Another result of

a lower frequency or long wavelength is the greater size requirement for an antenna

[19, 12]. Depending on the environment and physical space this may be a limiting

factor.

Millimeter Wave Radar

Finally, we come to millimeter wave (mm-wave) radar. Again, it is its wavelength

and frequency that differentiates it from laser and microwave radar. It falls between

microwaves and visible light on the EM spectrum with a frequency between approxi-

mately 30 and 300 GHz and a wavelength ranging from 1 to 10 mm, hence the name,

millimeter wave radar. Due to the shorter wavelength, the mm-wave has a nar-

rower beam width than microwave radar which makes a higher resolution and further

range possible. Additionally with a shorter wavelength, the component and antenna

3.3. CIGAR LAKE SENSOR SELECTION 24

size can be smaller. Unfortunately, small size requires precision manufacturing, and

hence, a correspondingly high cost. Availability of such a device is also limited and

procurement would be a potential problem.

Radar is an attractive technology due to its ability to penetrate dust, smoke, and

fog and thus its range imaging capabilities have been examined for use in adverse

environmental conditions [14, 51, 19], and specifically for mining applications [39, 11,

52, 64, 12]. Of course, the specific application, and its corresponding requirements

for cost, size, accuracy will drive the selection for any case.

3.3 Cigar Lake Sensor Selection

The process of sensor selection for Cigar Lake is discussed in the following section.

3.3.1 Historical Options Analysis

With the challenging environment of the jetted cavity in which the range measurement

system must operate in, the task of finding an appropriate sensor is not a simple

endeavour. Based on the experience gained in field testing, as discussed in Section

2.1.2, it was determined that an ultrasonic sensor would be used for the interim survey

and that a laser scanning system would be used for the post-cavity survey. However,

given the time between the initial testing in 2000 and expected start of production

(post 2008), there was opportunity to explore advances in technology and examine

further options.

In 2005, the Saskatchewan Research Council (SRC) conducted a review of cur-

rently available technologies for use in cavity surveying [37]. Several options were

considered for the post cavity survey, but only one showed significant potential based


on the restrictions imposed from the environment and deployment method. This was

the MDL Cavity Auto-Scanning Laser (C-ALS) system. It had been developed

specifically for cavity surveying applications and came complete with data processing

software. Upon a re-evaluation conducted by Cameco in 2012, following the McArthur

River deployment testing (see Section 2.2), the C-ALS was found to meet the size

and weight requirements for deployment using the rod pusher through the backfill

pipe. The C-ALS probe diameter is 50 mm, easily fitting within the 127 mm back-

fill pipe, and weighs 5.9 kg. The weight of the attached power and data cable were

also considered. Instead of using the standard cable, which weighs in at 0.18 kg/m,

a custom option was chosen with a lesser weight of 0.065 kg/m. This was to ensure

a deployment distance of 60 m could be easily achieved even after adapters, which

had yet to be fabricated, were attached. It is clear how all components, including

adapters and cables, must be considered as part of the restrictions for weight and

size of a post-cavity survey tool. However, even with the basic specifications met,

the system was expensive and its performance in the challenging Cigar Lake cavity

application had been not been validated. As discussed in Section 2.1.2, a previously

tested laser system had shown to experience issue with the water and fog environment

created by the jetting process.

In 2006, SRC carried out performance testing for three ultrasonic sensors to be

used in the interim cavity survey, including the Senix TX-30S1-ISR, the Massa

M-5000/95, and the Omega LVU-301 [43]. They were chosen based on potential to

withstand the extreme environment, and were characterized according to the following

parameters:

power requirements,


voltage sensitivity,

water effects

resolution at varying distances,

response to varying surfaces (texture, bends, spikes, and fissure),

response and sweep time,

reflection characteristics,

vibration resistance, and

screen protector tests.

Based on the results obtained from all the sensors, it was determined that the Senix

was the best choice since it had good accuracy, high response time, NEMA-6P rating,

stainless steel housing, no failures during testing, and had the smallest sensor size.

An alternative technology that SRC had also investigated for the interim cavity

survey application was the Vega Radar Sensor, which uses microwave radar. It was

found that it would be possible to modify an instrument, but that the compromises

necessary may not give an advantage over the ultrasonic units already being consid-

ered. It would require waterproofing and the antenna horn would have to be cut to fit

size requirements, resulting in an increase in beam width [37]. For this reason, their

focus remained on an ultrasonic sensor, with low cost, wide availability, and proven

capabilities in a fog environment, for providing range between periods of jetting for

the interim survey system.

Based on the signal properties discussed in Section 3.2.3, with respect to fog

penetration and a narrow beam width, the millimeter wave radar would seem well


suited to the underground cavity surveying environment. Graham Brooker et al.

[12] developed a 94 GHz FMCW mm-wave radar system for implementation in a

mining environment and found that, in comparison to sonar, laser, and microwave

radar systems, the mm-wave radar offered the best performance within a dusty or

humid environment and could withstand underground and surface mining operations.

Julian Ryde and Nick Hillier compared two laser range finders, a SICK LMS291-S05

and a Riegl LMSQ120, to a two-dimensional (2D) HSS 95 GHz scanning mm-wave

radar provided by the Australian Centre for Field Robotics (ACFR), using a test

chamber and in the field on an electric rope shovel [51]. They found that neither

sensor could alone provide sufficient data in the adverse environment, suggesting

that radar returns could be used to provide a rough draft of the surrounds, when

adverse weather and dust were present, but that a laser would be needed for detailed

information used in volume estimation and object classification. Castro and Peynot

[14] also examine using the combination of laser and radar for a perception system in

adverse outdoor environmental conditions on an unmanned ground vehicle (UGV).

The need for using two range sensing modalities had already been recognized for the

Cigar Lake application, though, instead of mm-wave radar, an ultrasonic sensor was

planned for use in the fog environment. Due to the high cost and limited availability

of mm-wave radar, it would not be a practical alternative.

3.3.2 Time-of-Flight Camera

The time-of-flight camera is a device which has been gaining increased use in research

for 3D imaging applications, such as map building [42, 34, 54, 57] or object recognition

[23, 21, 24], but had not been examined for remote cavity surveying. The Institut


de Robotica i Informatica Industrial (IRI) cited several examples of ToF camera

application in a technical report [22], and concluded that the devices most exploited

feature is the capability to provide complete scene depth maps at high frame rate

without the need of moving parts. In a survey of various ToF cameras currently on

the market, it was suggested that they will replace previous solutions, or at least

complement other technologies, in many areas of application [20]. The potential time

savings, sensor robustness, and small physical size that the TOF camera possessed,

provoked further examination for the Cigar Lake application. The basic operational

principles and potential advantages of using this technology over a laser system, such

as the C-ALS, is discussed within this section and constitutes the focus of this

dissertation.

The MESA Imaging SwissRanger (SR4030) was chosen for the Cigar Lake appli-

cation due to its small size, weight, and commercial availability1. It has dimensions

of 65.40 67.40 76 mm, and weighs a mere 410 g. The camera would easily fitinside the backfill pipe and could be deployed using the vertical rod pusher. The

detection range for the chosen SwissRanger camera was 10.0 m, and since the target

cavities had a diameter of 4.5 m and height of 6 m, it would easily meet requirements

for cavity surveying purposes.

It uses a CCD/CMOS imaging sensor where a continuously-modulated infrared

signal is emitted, reflected by the objects in the scene, and the precise time of return

measured independently by the sensor pixels inside the camera. Each pixel on the

sensor demodulates the incoming light signal and recovers the sine wave function from

which the phase delay of the recovered signal can be used to calculate target distance.

From the single pixel values on the imaging sensor, a 176 144 pixel depth map is1See http://mesa-imaging.ch.


computed.

In a report published by Swiss Center for Electronics and Microtechnology (CSEM)

[41], a mathematical model for the SwissRanger ToF camera is detailed and replicated

below for completeness. A graphical representation of the modulated signal is shown

in Figure 3.3.

The emitted signal, e(t), can be approximated as:

e(t) = e [1 + sin

(F

2pi t)]

(3.4)

and the reflected signal, s(t), as:

s(t) = BG(t) + e k [1 + sin

(F

2pi t

)](3.5)

where F is the modulation frequency in Hz, e is the emitted mean power in W, BG(t)

is the background illumination power in W, k is the attenuation factor including

target (distance, reflectivity) as well as the optics (lens, filter) and, is the phase

delay arising from the objects distance.

With the reflected signal being sampled four times in each cycle, at four period

phase shifts (i.e., 90 phase angle) it is possible to obtain:

= arctan

(A3 A1A0 A2

)(3.6)

B =A0 + A1 + A2 + A3

4(3.7)

A =

[A3 A1]2 + [A0 A2]2

2(3.8)


where is the measured phase delay, B is the measured offset, and A is the measured

amplitude.

The offset, B, represents the conventional black-and-white image and the ampli-

tude, A, is a measure of the quality of the acquired distance information. According

to the SwissRanger manual [5], the amplitude is converted into a value which is

independent of distance and position in the image array, using the following factors:

1. A factor proportional to the square of the measured distance, scaled to equal 1

at a distance of half of the full-phase distance.

2. A factor which corrects for the drop in strength of the illumination away from

the center of the filed of view. This factor equals 1 at the center, and increases

with radial distance from the center.

With the measured phase delay, it is possible to directly calculate the distance

from the target object to the camera, as shown below:

L =L02pi , L0 = c

2f(3.9)

where L is the target distance in m, L0 is the distance in m corresponding to one full

cycle, f is the modulation frequency in Hz, and c is the speed of light in m/s.

3.3.3 Device Comparison

As shown in Table 3.1, the SwissRanger device is the fastest device, obtaining up to

50 frames of data per second, where the C-ALS obtains only 250 data points per

second (see Appendix B for device specification sheets). The data from the C-ALS

is plotted in real time but would take over a minute to collect and display what the


Figure 3.3: Time of Flight sampling of returned modulated signal [5].

Figure 3.4: Devices compared for Cigar Lake cavity surveying. From left to right:MESA SwissRanger (ToF Camera), MDL C-ALS (Laser ScanningTool), Senix ToughSonic (Ultrasonic Sensor).

SwissRanger can obtain in 1/50-th of a second. The dense data set makes it possible

to employ advanced data analysis techniques such as registration or segmentation

as described in Section 3.4. It was also proposed that with such a dense data set,

it may be possible to filter spurious readings expected in foggy conditions [18], and

examine whether the resulting output could be interpreted. In this case, it would be

possible to use the ToF camera sooner than the C-ALS in foggy conditions and as

a supplement or even replacement for the ultrasonic sensor. Clearly, any time savings

3.4. POINT CLOUD PROCESSING 32

in a production driven environment would be seen as a cost benefit.

In addition to the dense point cloud data, the SwissRanger also provides amplitude

data. Since the amplitude data can be viewed as a grayscale image, this characteristic

would offer a visual representation of the target surface. The C-ALS is equipped

with a video camera, but it is mounted at the end of the probe. Its purpose is limited

to visualization during deployment through a conduit (such as the JBS backfill pipe),

as it does not provide the resolution or range sufficient to view the inside of a cavity.

With regards to the mechanical characteristics of the sensors, unlike the Swiss-

Ranger, laser systems generally employ several moving parts. Often with increased

complexity, comes increased maintenance. It is known that in ideal conditions, laser

systems have outperformed TOF cameras [16] for the purpose of very detailed 3D

imaging, but for the Cigar Lake application, the TOF camera could prove to be more

reliable. A distinct advantage of the C-ALS for a commercial application is that it is

a complete off-the-shelf system, equipped with application software, integrated pitch

and roll sensors, and a deployment method. To form a complete system with the ToF

camera, it would require system development, including adaptors for deployment, en-

coders, and software, along with required electronics and power. One contribution

of the research presented by this thesis is to evaluate whether the data acquired by

using the SwissRanger may offer advantages warranting further development for the

Cigar Lake application.

3.4 Point Cloud Processing

As discussed in Section 3.3.3, one advantageous characteristic of the ToF camera is

the substantially large amount of data that can be acquired in a very short amount


Table 3.1: Comparison of survey device specifications.

Senix ToughSonic MDL MKIII MESA SR4030TSPC-30S1 C-ALS ToF Camera

Survey Application Interim Cavity Post Cavity UnknownCost Low High MediumData Acquisition Rate 20 points/s 250 points/s 50 frames/sGrayscale Image No B&W video YesPackaging IP68 IP67 IP67Fog/Water Conditions Good Poor UnknownClosed Space Unknown Good GoodSurface Reflectivity Unaffected Unknown UnknownRange Accuracy 0.2 % of range 5 cm 15 mmDevelopment Stage Custom Commercial Required

of time. This opens up multiple options for analysis. Two options, registration and

segmentation, were identified to have potential for use in the Cigar Lake Cavity survey

application and are discussed in this section.

3.4.1 Registration

Ultimately, it will be necessary to create a full 3D model of a surveyed cavity by

piecing together the frames of data collected by a ToF camera. The process of con-

sistently aligning the frames of data is known as registration. Two overlapping views

of a surface are considered to be registered if a single transformation is found that

will bring one image into the correct pose within the coordinate system of the other

[15]. A survey system such as the C-ALS uses an encoder that tracks the relative

position of the laser head to form a 3D model. An encoder could similarly be used

to provide the ToF camera viewpoint, however, it is proposed that registration could

further refine the image alignment. The combination of multiple modes to determine

pose, would be of particular use in this remote application, where there are limited


options for data validation.

One of the most common methods for registering point clouds is by use of the

Iterative Closest Point (ICP) algorithm [9, 65]. The widely-used ICP algorithm func-

tions through a process of iteratively refining the estimated rotation and translation

between two frames of point cloud data by minimizing the mean-square distance be-

tween their points [9]. Due to its popularity, there exist many ICP variants [47], so for

a base case, example code2 from the Point Cloud Library (PCL) was proposed for use

with the SwissRanger. PCL is a large scale, open project for 2D and 3D image and

point cloud processing3 that contains many advanced algorithms for filtering, feature

estimation, registration, segmentation, and more [50].

The example pipeline for PCL ICP4 includes the following steps:

1. Search for point correspondences

2. Reject bad correspondences

3. Estimate a transformation using the good correspondences

4. Iterate at step 1

In Section 4.5 of this thesis, the results of registration attempts using data from

the test cavity are presented.

3.4.2 Segmentation

In order to further process the large amount of data from dense point clouds, or to

extract additional information, methods of segmentation can be used. Segmentation

2See http://pointclouds.org/documentation/tutorials/pairwise_incremental_registration.php.

3See http://pointclouds.org/.4See http://pointclouds.org/documentation/tutorials/registration_api.php.


is the process by which data points of similar properties are grouped or clustered

together. It is possible to look for object or surface edges, identify different surface

textures and colours, or cluster the data according to whatever parameter will serve

a particular purpose [48, 29, 44].

To achieve maximal efficiency in mining at Cigar Lake, it is important to extract

all targeted ore while minimizing waste or dilution for each planned cavity. Currently

the mine plan is based on core samples that were used to develop a block model. If

it were possible to segment the boundaries between waste rock and the ore deposit

while the jetting tools were in-hole, benefits could be realized during the interim

survey, ensuring that jetting process extracts the targeted ore, or even afterwards, in

validating the block model and planning future cavities.

The most obvious differentiating property between waste rock and the uranium

ore deposit is their colour. Referring to Figure 4.16(a), the red-brown coloured core

is hematized clay, and the gray colour would be either choritized clay or the gray-

black chloritized sandstone. The particular area of interest is the very black coloured

mineral which comprises the pitchblende U3O8 deposit. If the reflectivity of the

different minerals were found to be distinct, it would be possible to segment the

SwissRanger amplitude image accordingly and map the ore body extents during or

after mining is complete.

In Section 4.5 of this thesis, segmentation is tried on data collected from inside

the test cavity.

36

Chapter 4

Experimental Studies

4.1 Test Cavity

Device testing in an experimental test cavity was conducted to demonstrate the ca-

pabilities and limitations of the devices proposed for use within the underground

cavity application at Cigar Lake, as compared with the ToF camera. The tests were

intended to provide a baseline for the interpretation of data obtained from any of

the systems in a real remote cavity and to allow for comparison between devices.

Since the shape, size, and target reflectivity of the cavity have an effect on the sensor

performance, a testing space was designed to emulate the properties and size of a

true cavity as closely as possible. A hexagon-shaped wooden enclosure was built with

wall-to-wall distances varying from 4.2 m to 5.0 m. A wall shape similar to a cavity

was constructed with stucco. Diamond mesh was manipulated to cover the surface of

the walls and a stucco base was applied. To finish, Cigar Lake core samples were ex-

amined and used as a guide to select colours for the stucco finish (see Figure 4.16(a)).

The uranium ore is very strong and likely to protrude further than the surrounding,

softer rock, when jetting is conducted. The high grade ore is also pitch black and for

4.1. TEST CAVITY 37

this reason, the greatest protrusions on the test cavity walls were painted black (see

Figure 4.1).

(a) A real Cigar Lake cavity with freeze pipes (b) Test cavity construction

(c) Completed test cavity

Figure 4.1: Design and construction of test cavity.

Testing first included baseline data acquisition, where effects (if any) due to the

closed space and uneven surface composed of various reflectivities could be observed.

To follow, the application specific environment was simulated and the effect of fog

and water on the range data, as well as the detection of freeze pipes, was evaluated.

4.2. TEST APPARATUS 38

The quality of the data obtained, the time in which it could be acquired, the asso-

ciated cost, and the JBS operators ability to interpret the results all play a role in

determining which sensor is best suited for the interim and post cavity surveys.

4.2 Test Apparatus

The C-ALS is part of a complete scanning system but required stabilization within

the test cavity to complete experiments. A bike repair stand was found to be sufficient

for this purpose and the setup is shown in Figure 4.2. The C-ALS possesses an

actuated head and so, once the system was appropriately positioned, a 3D scan could

be initiated with the use of the application specific software.

Figure 4.2: C-ALS test apparatus in test cavity.

Since the SwissRanger ToF Camera is a stand-alone sensor, a Celestron NexStar

SE tripod was used to rotate the camera and output the angle of rotation. The

ultrasonic sensor and SwissRanger were mounted to an adaptor designed by SRC, as

shown in Figure 4.1(c). Scripts were written in MATLAB (see Appendix A) and

4.2. TEST APPARATUS 39

used to trigger data acquisition from the tripod, Senix sensor, and SwissRanger.

Basic scripts were also used for preliminary data viewing and analysis.

With the need of an efficient means to later process the large amount of data that

was obtained from the SwissRanger, the Point Cloud Library (PCL) was used. Inside

PCL, the Point Cloud Data (.pcd) file format is used, which requires that a specific

header be used to declare certain properties of the point cloud data stored in the file.

As a result, a script was written in MATLAB to save the data using the PCD file

format. The tripod rotation angle was converted into quaternions, representing the

cameras viewpoint, for the PCL header file. The use of quaternions provides a useful

way of performing rotational mathematics in 3D space [17], and are utilized in the

PCL viewer. The unit quaternion q= [q, qx, qy, qz] is computed as

qw = cos(/2) (4.1)

qx = i(x sin(/2)) (4.2)

qy = j(y sin(/2)) (4.3)

qz = k(z sin(/2)) (4.4)

where the vector R3 is the axis of rotation, R represents the angle of rotationabout , and i, j, and k are unit vectors on the x, y, and z axes, respectively.

Since the motion took place in the x z plane of the camera (see Figure 4.3),with rotation about the y-axis, a simplification was possible whereby = [0, y, 0].

As a unit quaternion, with the condition that |||| = 1, this would further simplifyto = [0, 1, 0].

4.3. TEST ENVIRONMENTS AND RESULTS 40

Figure 4.3: SwissRanger camera Cartesian coordinate system, (x, y, z) [5].

4.3 Test Environments and Results

4.3.1 Baseline

Before beginning the series of various experiments, it was important to determine the

appropriate settings for each device and to acquire a set of baseline data for each.

The rationale behind the base settings and those that would be varied for each device

is presented within this section.

Senix

Range data was obtained with the Senix ultrasonic sensor at the same position that

an image was also acquired with the SwissRanger. It was regularly found that the

Senix sensor would provide a false range of 0 m, likely dependent on the particular

angle of incidence to the surface. For the interim cavity survey system developed

by SRC, this was remedied by filtering out the 0 values and creating an algorithm

to smooth the remaining range data on a continual basis as the sensor was rotated.


The value of further analyzing data from the Senix sensor was deemed unnecessary

for the purpose of this dissertation, since its primary advantage had already been

recognized in the capability to provide range measurement in fog conditions.

C-ALS

MDL supplies a software package with the C-ALS that allows for a few options

in the acquisition process. The first considered was a high speed option (default)

which increases the speed of a survey by 20 % but provides 20 % less data. Since

the quantity of data was deemed sufficient, the high speed option was selected for all

testing. Another consideration was the Last Hit option which is intended for use if

there is an obstruction between the instrument and the target, such as water vapour

or dust [35]. These obstructions may cause the laser to reflect back to the instrument

before reaching the intended target. Testing was completed with this option set both

on and off to observe and compare the effect to data acquired in fog (see Section

4.3.4).

The final scanning option reviewed was the interval angle for either vertical or

horizontal slices of data that could be acquired. Figure 4.5 shows how the interval

angle affects the visual detail of the 3D plots. There is an obvious improvement in

the visual information provided between a scan taken with a 5 interval to that of 1.

However, the difference between 1 and 0.5 appears less obvious, even though there

are approximately twice as many data points and the scan will have taken twice as

much time (see Figure 4.4). This will be a consideration for acquiring data in the

field where time has a direct correlation with cost. It is necessary to collect sufficient

data for cavity modelling and mine planning, but in a minimal amount of time. For


data sets collected in the test cavity, the interval angle was set at 1, since time was

not of essence (in contrast to the mine), and the data set could be down-sampled, if

necessary.

Figure 4.4: C-ALS baseline scan (vertical) demonstrating data acquisition time.

SwissRanger

With the SwissRanger, it is possible to acquire four basic sets of data (see Figure

4.6), which include:

1. 3D Cartesian coordinate data (x, y, z) (point cloud),

2. range image,

3. amplitude (grayscale) image, and

4. confidence map.


(a) 5 interval

(b) 1 interval

(c) 0.5 interval

Figure 4.5: C-ALS vertical scan 3D plots shown with increasing acquisition intervals(colour scaled by signal strength with blue for low and red for high).


These basic sets were acquired for each triggered data acquisition.

Four different filtering modes, intended to reduce noisy data, could be set on the

SwissRanger and include:

1. Raw Data - No filters

2. Median Filter - 3 3 median filter run on the host PC

3. Neighborhood Filter - 5 5 hardware adaptive neighborhood filter

4. Median and Neighborhood Filter

Data was collected with each filtering mode, but for consistency, the setting for

Median and Neighborhood Filter was most often used in analysis.

The integration time, which is the length of time that the pixels are allowed to

collect light, was also varied during data acquisition to observe the effect within the

test cavity as a baseline and also in the different test environments. The parameter

intT ime was set through the SwissRanger API and is related to integration time, IT ,

the read out time, RO, and the cameras frame rate, FR, by the following equations:

IT = 0.300 ms + (intT ime) 0.100 ms (4.5)

FR =1

4 (IT +RO) (4.6)

As discussed in Section 3.3.2, four samples of the phase are needed to calculate

range, requiring four separate integration periods. Thus, the time required to capture

a single image frame is the inverse of Equation 4.6.


Figure 4.6: Left: Basic image acquisition from SwissRanger using MATLAB (ampli-tude (grayscale) image, range image, and confidence map). Right: Pointcloud data plotted using the PCL viewer.

As expected, it was observed that visual appearance of the target wall surface

in the amplitude and range image improved as the integration time increased, and

respectively, the noise appeared to decrease.

4.3.2 Water

For the Cigar Lake application, the presence of water in the cavity being surveyed

is expected. During the process of jetting, the nozzle produces a jet of water that is

directed upwards at an angle of 65 from the axis of the sub, with flow at the nozzle

at pressures of up to 100 MPa. The interim survey, taking place between periods of

jetting, will be completed with the Senix ultrasonic device, but it is also expected


that the cavity surface will remain water saturated for the final, post cavity survey.

Therefore, any ranging device to be used must operate reliably with water on the

target surface and potentially on the sensor itself.

In order to simulate the cavity environment after jetting, water was sprayed inside

the test cavity, saturating the walls and roof and creating near 100 % humidity within

the space. Data obtained from both the C-ALS and SwissRanger were compared to

the baseline data and found to have no significant difference in the range data, as

shown in Figure 4.7.

The water was also sprayed directly onto the sensor surfaces to observe what effect

it would have on performance. For all sensors, the effect was most noticeable directly

after spraying occurred because it caused a distortion of the signal. A shorter range

(by 10 cm) was measured with the Senix ultrasonic sensor and the data fromthe C-ALS appeared noisy with early signal returns. Noisy point cloud data was also

observed with the SwissRanger and the look of a lens appeared on the confidence

image (see Figure 4.8). The effect water had on each sensor diminished over time as

the heat dissipation from the sensors caused the water to pool and evaporate.

4.3.3 Freeze pipes

As part of the mining process at Cigar Lake, the ore and surrounding rock must be

frozen prior to jetting. This is achieved through the installation of freeze pipes in

a grid pattern, through which brine is circulated to freeze the ground. It is known

that jetting cavities will expose freeze pipes and it is beneficial to know how these

will be imaged by each device in order to identify them during the surveying process.

If a freezepipe is identified, it could be factored into the calculation for determining


Figure 4.7: SwissRanger Point Cloud data with water (blue) and without water(green) on test cavity surface.

the overall volume the cavity. Perhaps more importantly, a freeze pipe that has been

exposed will need to be identified so that it can be monitored when the brine is turned

back on, ensuring no permanent damage has occurred.

The ultrasonic device has a nominal beam width of 12 and is unlikely to resolve

a freeze pipe. With some basic calculations, considering the outer diameter of the

freeze pipe, it can be found that even if the beam were centered on the mid point of

the pipe, it would be improbable to resolve a pipe that is beyond a distance of 0.60


(a) Point cloud (b) Amplitude image

(c) Range image (d) Confidence image

Figure 4.8: SwissRanger Data with Water Droplet on Lens.

m from the sensor. Furthermore, the current software uses a smoothing average, so

it would be even more unlikely to distinguish a point reflected on a freezepipe from

the averaged points on the proximate cavity wall.

With the C-ALS, there are three typical scan interval settings. These are at 6,

3, and 1. It is clear that with the more data, you are more likely to detect the freeze

pipe, but consideration of time must also be a factor. It was observed that with a

vertical scan interval of 3, as shown in Figure 4.9, the freeze pipe could be identified

but without a high level of confidence.


(a) Front view

(b) Side view

Figure 4.9: C-ALS scan of freeze pipes. Scanning interval (Left to Right): 6, 3,and 1.

In contrast, with the high density of range data obtained in a single image with the

SwissRanger ToF camera and the corresponding amplitude image, as shown in Figure

4.10, the chance of a freeze pipe going undetected would be highly unlikely. Object

detection is clearly an advantage of the SwissRanger over the C-ALS, especially

with the contribution to visualization from video frame rate data acquisition.

4.3.4 Fog

As determined from testing in 2000 (Section 2.1.2), the build up of fog is a significant

issue for any survey sensor that is to be used in the jetted cavity where water is sprayed

in the frozen environment. To simulate fog conditions, two methods for creating fog


(a) Point cloud (b) Amplitude image

(c) Range image (d) Confidence image

Figure 4.10: SwissRanger data with freeze pipes in test cavity.

were assessed. The first method simply involved mixing boiling water with dry ice.

It was quickly determined that this was not a viable option to create a consistent fog,

since it sunk to the floor and dissipated quickly. Instead, an Antari fog machine

was used to create the suspended particulates or fog within the test cavity. Inside

the machine, a mixture of glycol and water is passed through a heat exchanger where

it is vapourized, forming a fog when mixed with the cooler air outside the machine.

Approximate light, medium, and dense fogs, as shown in Figure 4.11, were created by

timing the length of vapour release. A shortcoming of this experiment was the lack


of a specific fog density metric. To assess device performance in fog, permutations of

the filter settings were conducted for both the C-ALS and SwissRanger.

(a) Light fog (b) Medium fog

(c) Dense fog

Figure 4.11: Inside of test cavity with light, medium, and dense fog conditions.

The C-ALS Last Hit option, as mentioned in Section 4.3.1, can be used when

there is water vapour or dust in the air. This option, however, did not make an

observable improvement on the data obtained within the test cavity. It is possible

that even the light fog density was too great for a sufficient amount of the signal

to reach the target surface before being returned. In comparing the baseline and


dense fog data from the C-ALS, it could be seen that the point cloud obtained in

fog still resembled the true surface but was scaled down, due to the early returns

of signals on the fog. The new points appeared roughly 1 m from the true target

surface. In application at the Cigar Lake mine, it will be important to identify when

fog is present in the cavity because false, early-returns may not be distinguishable

from normal data and therefore cause significant errors in volume and production

estimates. It would be difficult to assess the impact of the fog on the C-ALS data,

and attempting to correlate fog density to the impact on sensor reading is beyond

the scope of these experiments. However, the importance of utilizing the integrated

video camera on the C-ALS to observe the conditions within the remote cavity and

validate the dissipation of fog, prior to data collection, was substantiated.

Figure 4.12: C-ALS data in dense fog (left) and baseline (right). Target distanceshown on horizontal axis in meters.


During testing with the SwissRanger, it was observed that the point cloud was also

scaled down according to the level of fog inside the test cavity (see Figure 4.13). The

average range for the baseline data was 2.00 m and for the point cloud in dense fog

was 0.21 m. It is clear that the lesser power of the SwissRanger signal, as compared

to the C-ALS, had a significant effect on signal penetration through the fog. Since

the 2D amplitude image of the cavity wall still resembled the true surface (see Figure

4.14), despite the small range values, it raised the question as to whether the accurate

range values could be extracted or filtered with a lesser integration time. It had been

noted that part of the point cloud data still encompassed true range values, with a

shorter integration time, though it was unknown whether this data was simply noise.

A simple setup was devised in order to determine whether it was feasible to extract

true range data from the SwissRanger in fog. The camera was located at two different

positionsapproximately 2 m and 4 m from the target surface within the test cavity

and data taken at several integration times, in the light, medium, and dense fog

conditions. It was expected that noise would appear the same, indifferent to the

camera distance from the target surface and the level of fog. As shown in Figure

4.15, at an integration time of 0.3 ms (see Equation 4.5), the point cloud consists

purely of noise, since there is little difference observed at both positions and between

the baseline (no fog) and medium fog conditions. At a higher integration time of

5.3 ms, the baseline data still does not represent the true range, but it is possible

to see the point clouds from the two different positions begin to separate. With the

same integration time, but in fog, the point cloud begins being condensed for both

positions. Finally, at the highest integration time of 25.3 ms, the true average range

is represented at approximately 2 and 4 meters without fog in the cavity. With fog,


0.51

1.52

0.5

0

0.5

1

0.5

0

0.5

Range (m)Width (m)

Heigh

t (m

)BaselineLight FogMedium FogDense Fog

Figure 4.13: Point cloud data from SwissRanger in light, medium, and dense fogconditions inside the test cavity.

it is exclusively early returns that are observed, thus demonstrating the difficulty in

extracting the true range. It was the highest integration time that was required to

achieve reliable range data in clear conditions, but it was not feasible to obtain the

same within fog.

In the field, the possibility of extracting true range data in a fog environment

would diminish further. The density and composition of fog would be variable and

the target distance would certainly not remain the same as jetting progressed. For


Pixel Array Width (No.)

Pixe

l Arra

y Heig

ht (N

o.)

20 40 60 80 100 120 140 160

20

40

60

80

100

120

140 100

200

300

400

500

600

700

800

900

1000

(a) Baseline amplitude image


Pixe

l Arra

y Heig

ht (N

o.)

20 40 60 80 100 120 140 160

20

40

60

80

100

120

1400

20

40

60

80

100

120

140

160

180

(b) Amplitude image in medium fog


Pixe

l Arra

y Heig

ht (N

o.)

20 40 60 80 100 120 140 160

20

40

60

80

100

120

1400

10

20

30

40

50

60

(c) Amplitude image in dense fog

Figure 4.14: Amplitude images from SwissRanger in fog (Auto-scaled inMATLAB).

the remote cavity application, it is necessary that a certain level of confidence in the

range data could be achieved and this was not seen to be possible with the C-ALS

and SwissRanger. The C-ALS possessed a stronger powered signal and thus was

able to acquire range data closer to the baseline than the SwissRanger, however, it

was observed that both devices would provide an inaccurate indication of the size and

volume of the cavity in fog conditions.

4.4. FIELD TESTING 56

0 1 2 3 4 5 6 7 8 9 104

3

2

1

0

1

2

3

4

Baseline DataIntegration Time = 0.3 ms

Range (m)

Heigh

t (m

)

0 1 2 3 4 5 6 7 8 9 104

3

2

1

0

1

2

3

4

Baseline DataIntegration Time = 5.3 ms

Range (m)

Heigh

t (m

)

0 1 2 3 4 5 6 7 8 9 104

3

2

1

0

1

2

3

4

Baseline DataIntegration Ti

Remote Surveying of Underground Cavities.pdf

Documents

Transcript of Remote Surveying of Underground Cavities.pdf