PHOTOORAWETRY SOFTWARE. A PACKAGE FOR EVERYONE,(U)OCT 81 J R HAWK

UNCLASSIFIED NL

lllll omoil.!i!!iFiiI'E l"IA1"'

11111 *~ 136

138

1111111 111111.4____ 1J.6

MICROCOPY RESOLUTION TEST CHARTNATIONAL BUREAU OT SIANDAR9S 9ME A.

SECURITY CLASSIFICATION OF THIS PAGE (We ots Entered)

' REPORT DOCUMENTATION PAGE BFO = OM

I. REPORT NUMBER 2. GOVT ACCESSION .QW RECIPIENICS T N!F

U4. TITLE (and Subtitle) S YEO EOTPRO OEE

Photogrammetry Software N/IPA Package for Everyone _______________

6. PERFORMING ORG. REPORT NUMBER

N/A7. AU THOR(o) 8. CONTRACT OR GRANT NUMBER(&)

James R. Hawk N/

PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT PROJECT. TASKAREA & WORK UNIT NUMBERS

['MA Inter American Geodetic SurveyN/Bldg 605N/CO Fort Sam Houston, Texas 78234 ______________

CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE

.DMA Inter American Geodetic Survey October 1981* Wh~ Bldg 605 13. NUMBER OF PAGES peni

)rt Sam Houston, Texas 78234 16 in Report, 3A in Apni

MONITORING AGENCY NAME & AOORESS(f different from Controlling Office) IS. SECURITY CLASS. (of this report)

-~~ Unclassified i

I5s. DECLASSIFICATION/DOWNGRADING XSCHEDULE N/A

SApproved for public release; distribution unlimited.

N/A

Ifl. SUPPLEMENTARY NOTES

To he presented and published at the 2nd Technology FxhneWeek, Japq~ry1982, Panama City, Panama.To he submitted to the PAIGH Conference, "arch 19P2, Santiano, Chile.

19. KEY WORDS (Continue ont reverse olde, it necessry mid identity by btock number)

SW!'A Adjustment ProgramGIAMT Adjustment Program ''Y''? -

Panoramic photography D .. i~IO

S21L AftrRACT I'Contilm -f reverse ebbe If neoweiv Mill Ideuttlfy by block nimabor)

An example photogrammetry software system is presented for consideration.The system is being implemented throughout Latin America by TAGS. It includesboth analytical and semi-analytical adjustments. It is a simplistic yet ver-

Wsatile system which has proven very successful.4,

D F 0=03M EDITIOw OF I NOV 65 IS NO.104E. RrLSIF

SECUmITV CLASSIFICATION OF THIS PAGE (When Data ffntered)

'- - , i. -

DISCLAIMER NOTICE

THIS DOCUMENT IS BEST QUALITYPRACTICABLE. THE COPY FURNISHEDTO DTIC CONTAINED A SIGNIFICANTNUMBER OF PAGES WHICH DO NOTREPRODUCE LEGIBLY.

I

1~** - - - .-~.-. -....- -

PHOTOGRAMMETRY SOFTWAREA PACKAGE FOR EVERYONE

James R. HawkDefense Mapping Agency Inter American Geodetic Survey

Building 605, Fort Sam Houston, Texas 78234

BIOGRAPHICAL SKETCII

James Hawk is a photogrammetrist with the Defense Mapping

Agency Inter American Geodetic Survey (DMA lAGS). His

responsibilities include providing computer software sup-

port to the Cartographic School in the R.public of Panama

and to associate Latin American mapping agencies. Mr. Hawk

also has over ten years working experience with the DefenseMapping Agency Aerospace Center in St. Louis, Missouri.

He received his B.S. degree in mathematics and geography

from Indiana State University and his M.S. degree in photo-grammetry from Purdue University. Mr. Hawk is a member

and past regional officer of A.S.P.

ABSTRACT

The development of a photogrammetry software package

requires considerable study. Computer design and capacity,available photogrammetric instruments, analytical methods

versus semi-analytical methods, simplicity versus compre-

hensiveness, user competence, and the desired final productsare but a few of the elements to be weighed. An example

system is presented for consideration in this paper. Thissystem is being implemented throughout Latin America by the

Inter American Geodetic Survey. The package consists of

both an analytical and a semi-analytical adjustment program

and the accompanying programs which tie the systems together

and allow the user to go from relative orientation or com-

parator measurements to plotter element settings via computer.Included in the discussion is control selection and distri-

bution, pass point selection, program execution, and analysis

of results. The sample software package was assembled

because it is simpltmW and yet versatile enough to be

adapted to a variety of computers and photogrammetric equip-

ment. It is presented as a successful example of a system

design, and it is hoped that this discussion and example

will generate new ideas in system development.

INTRODUCTION

Photogrammetry may be defined simply as a method of extract-

ing information about an object by studying its image.Photographs are the most familiar types of images, but

devices sensitive to other parts of the electromagnetic

spectrum may also be used. Regardless of the type of sen-

sor, the image must be analyzed before the correct informa-

tion is obtained. The desired results could be interpretive

and require the art of photo-interpretation to identify andevaluate the image. However, the desired results could be

quantitative and require measurements and mathematics to

arrive at the desired goal.

DTI

,.s .'it.," t . This document bas been pproved ELECTE

or public telease and sale; its S3 1981

A~~i t Pributi* daO '

The advent of the computer and its adoption to photogram-metric problems has greatly enhanced this science. Evenin photo-interpretation the computer is used to recognizepatterns in even the nonvisible region of the spectrum.Quantitative methods have benefited immensely from com-puter utilization. It is now possible to solve thousandsof simultaneous equations in only a short time.

The computer software discussed in this paper was designed

to help obtain quantitative results from aerial photography.The results are used to obtain dimensions and precisepositions. The desire for increased accuracy has led tothe development of precise equipment, systematic procedures,and rigorous computer adjustments. The most common appli-cation of the resultant information is in the preparationof topographic maps and charts of the earth's surface.Almost all of the national mapping programs in the WesternHemisphere are performed by photogrammetry.

The goal of the software system developed by IAGS is toencompass the needs of all the associate agencies. It isnot burdened with copyright laws and therefore allowseveryone free access. It maintains simplicity and yetallows-us to obtain meaningful results from a wide varietyof equipment. Technical advancement in computer hardware,photogrammetric instruments, and mathematical formulationwill not allow any software system to rest. Our programs

are under constant change, and we are always looking forsomething better. The package, however, has been installedand is being successfully used in the United States and inseveral Latin American countries.

SYSTEM DESIGN

The design of any photogrammetric software package mustbegin with the adjustment program. It is the largest and

most complex part of the system, and the other programs Iare built around it. The wide variety of instrumentation,technology, and requirements found in the hemisphere pre-clude the use of a single program to accomplish this task.Thus both semi-analytical and fully analytical approacheswere taken. Both applications produce accuracy and effi-ciency sufficient to fulfill our mapping requirements.

The purpose of the adjustment program is to enhance theground control network that currently exists through fieldsurveying, doppler positioning, or other techniques. Eachmodel used in a stereoplotter requires at least two hori-zontal control points for scaling and three vertical controlpoints for leveling. It would not be economical to estab-

lish all this control by ground methods. Adjustment programsmust also provide a method of detecting blunders in theobservations, evenly distribute small systematic errors,

and determine estimates of precision for the results.

After the adjustment programs were selected, a series ofauxiliary programs were acquired and written to blend thetwo main programs into a versatile smooth-flowing system.

These programs allow us to check obiervation!; for blunders,check data dimensions so as not to exceed the adjustmentparameters, produce independent model coordinators fromtwo dimensional comparator measurements, mathematicallycompute projection centers, determine stereoplotter orien-

tation settings after the adjustment, and provide otheruseful information as will be discussed later in this

paper.

A depiction of the overall system flow is shown in Figure I.Raw data may enter the system from three different sources,

be channeled in various directions and adjusted as desired.

TIE SIMBA ADJUSTMENT SYSTEM

The Simultaneous Block Adjustment of Models (SIMBA) program

was written by Randle W. Olsen of the U.S. Geological Surveyin 1973. It has been adapted for use by the U.S. Forest

Service and DMA lAGS. It is written in FORTRAN IV and isused to simultaneously adjust independent model units to

each other and to ground control. The three-dimensional

data may be generated on a stereoplotter or on a comparatorproviding the CRAST (Coordinate Refinement-Analytical Strip

Triangulation) or a similar program is used to derive the

projection centers and the third dimension. The models are

adjusted by an interactive series of planimetry-height solu-tions. Four-parameter linear transformations are used inthe horizontal solution and three-parameter linear trans-fromations are used in the vertical solution. This separation

economizes computer time and storage without a significant

compromise in accuracy as compared to more traditional seven-

paramater adjustments.

Program features include:

1. Input arranged in model units with a separate control

deck.

2. Internal sorting of control and tie points.

3. In-core solution of banded normal equations to minimize

computer time.

4. Separate weighting for ground control and model ties.

5. Printed output arranged in model units with corres-

ponding residuals.

6. A listing of test point coordinates and residuals which

were withheld from the solution.

7. A root mean square (RMS) error summary of held control,

tie points, and unheld control.

8. Absolute orientation values to include common tilt,

common tip, the airbase, and exposure station elevationdifference for each model.

ti

I .VS PH1OTOG RAtMMl-T RY P ROG RAH FLOWCH[ART

Raw Data Raw Data Raw DataK

Stereo MIono Stereo

Comparator Compatator Plotter

Yes Disk PROCEN

omparato Program

Card SIMBA

ReformatSORT

PProgram

Reflned ndepndenPlateMode

SIMBA~FIUR CoriaesOtu

Special attent ion should be given to the we ight ing systemas it affords us the opportuntt to separate errors result-in4 trom poor control Crom error:; due to relative orientationor bad model tie points. The equation for the expressionof the weight ratio is:

'2 S il

Sm 2 + Sc2

Where Sm = estimated unit coordinate standard deviationof thie model. points

Sc = estimated standard deviation of each controlpo i n t

One can easily see that the value of "W" will range fromzero to two. A value of one will allow equal weight tobe placed on the control value and the model ties. A

value close to two will constrain the control and allowthe models freedom to move. Likewise a value close tozero will constrain the models and allow the control toshift.

A minimum of two horizontal control points are requiredfor thoe planimetric adjustment, and three vertical controlpoints are required for the height adjustment. If no

additional control is used, there will be an absolutesolution and no errors will be propagated by a least

squares adjustment. Thus if only minimum control isused and constrained while the model ties are given free-dom to move, the propagated error will result from poormodel fit, and the tie point residuals will be exaggerated.Actual errors present in the constrained control are un-important because they are only being used to scale andlevel the block solution, and we are only checking tiepoint residuals.

Once a good model tie solution is achieved, the full control

network is used, but the weight is changed to constrain themodels and give freedom of movement to the control. The

control residuals thus become exaggerated, and controlvalues incompatible with the relative model solution canquickly be noted. After all blunders have been eliminated,equal weighting is generally used in the final solutionunless extenuating circumstances dictate otherwise.

The computer memory necessary to execute this adjustment islargely dependent upon the size of the normal equationcoefficient matrix. If the independent models are properlyaligned in the program, this matrix becomes banded along

the diagonal leaving only zeros outside the band. To con-serve storage, only the elements within the band above themain diagonal are stored. They are arranged in a rectangular

array called "Q" in the program and are dimensioned 4 x (thebandwidth) by 4 x (the number of models). Correct sequencing

of the models is therefore critical to stay within theallotted bandwidth. In a rectangular block, the band-width can roughly equal the number of models in the longeststrip plus two. The normal equations are solved by theGaussian elimination procedure using routines suited for

symmetric and banded equations. rhe banded routine con-serv'es computer core and s;ign ii icant ly reduces cxecut iont ime.

It is always difficult to recommend a control point patternbecause thare are so many variables, the most important otwhich is quality. Also, it has been my belief that a goodphotogrammetrist will attempt to secure as much controlinformation as possible, and his only limiting factor willbe the man in charge of expenditures. At the risk ofvioLating this rule, a control pattern which should satisfymost needs is recommended.

It is essential to have horizontal control in the extremecorners of the block. Any deviation from this will causewarping in the corners becoming progressively worse as thecontrol point strays from its corner position. Other thanthe corners, SIMBA seems to hold horizontal position ratherwell. Additional horizontal control should be spacedaround the edges of the block at six to eight model inter-vals.

Vertical control is a different matter. The adjustmentacts as a blanket tacked down at the corners with a fanblowing air under it. It bulges in the middle, and whenwe tack down the middle, we develop two smaller bulges oneach side. This progression continues geometrically untilthe bulges are reduced in size to an acceptable level. Thegeneral recommendation is to have a vertical control patternat three to four photo intervals throughout the block. Asample control point pattern is shown in Figure 2.

Pass and model point selection is a matter of personalcustom, but the author has developed a system which lendsitself well to both the adjustment mathematics and productionflow. Starting with one of the border strips in the block,pick four points in a line in all the trilap areas and atthe two ends. Insure that the two points closest to theadjacent strip fall in the sidelap area and are trans-ferred. Transfer these tie points to the second stripand send the first strip to the comparator or stereoplotterto be measured. Pick points on the second strip as before.If transferred points from the first strip happen to fallin the trilap or necessary end positions of the secondstrip, use them. If not, continue point selection as onstrip one. It is not necessary to transfer any of thenewly selected points back to the first strip. Continuein this manner until all points are selected. This methodwill produce from eight to ten pass points per model andis illustrated in Figure 3.

Three versions of SIMBA are currently available at DMA lAGSHeadquarters in San Antonio, Texas. These versions are thestandards which are then adapted to the computer and desiresof individual users. The first version is the most commonand has parameters of 240 models, 2200 unique points, 2500total points, 150 control points, and a bandwidth of 25.It uses 526K bytes of storage on an IBM 370 computer. Thesecond version of SIMBA breaks the program down into three

I"I

CONTROL PO[N' SELECT[O,'.

- 8 Models 3 - 4 Models

Vetia Coto

AIGR A

L A

II

D Horizontal Control

A Vertical Control

FIGURE 2

PASS AND TIE POINT SELECTION

C, io o ; ;o 00', 0 Strip 3

KM itbA !A - - J ,

Strip 2

iiiD 1 I,1 Aj Sidelap

A A &Z A A Strip 1

TrilapArea

Points selected on Strip I

C Points selected on Strip 2

0 Points selected on Strip 3

Between-strip tie points are shaded.

FIGURE 3

lrt whioh are executed ;eparatelV. Th t first part reads'i1 in itit Jata, org:1 nizes it, and buiiLds a fil . The

.cc, ond part reads that fi le, perf orms the necessary math,-

:uat io;, and writes the massaged data back onto the file.

The thi ird part reads the file and prints out the desired

reSults. This three part version is a computer core saver

but sacrifices time and simplicity. Dimensioned similar

to the original program, the three part version will save

about 75K bytes of storage. The third version we have isanother three part program with the first and third part!;

remaining as before. The difference in part two is that

a direct access disk file is used to write and then read

each row and column of the normal equation coefficient

matrix. All of this reading and writing takes time but

an additional reduction of 125K bytes of computer memory

is realized. This version is impractical unless the computer

memory is limited and computer time is of no concern.

A sample SIMBA output is shown in Appendix A.

The SIMBA adjustment requires the use of perspective centers

to strengthen the vertical solution. If a stereoplotter is

used for relative orientation, the coordinates for the per-

spective centers may be computed using either one of two

available programs. The ideal solution for determining

the camera perspective center is by space resection usinga reseau grid. The TAGS version of John McLaurin's Per-

spective Center Determination Program (U. S. GeologicalSurvey, 1969) does precisely this. A grid of known pre-

cision, usually engraved on a glass plate, is projected

through a stereoplotter projector. The coordinates offr--' three to fifty grid intersections are then measured

in model space. The bundle of rays passing through the

grid intersections on the calibrated plate and the bundleextending to the same measured intersections in model space

originate from the same point--the perspective center. If

the latter bundle of rays is fitted to the other bundle in

a least squares space-resection adjustment, accurate coordi-nates for the perspective center can be determined. Printed

output includes:

i. Title.

2. Input data to include point numbers, number of read-ings on each point and grid coordinates.

3. Number of points used and computed principle distance.

4. Mean projected grid coordinates and standard devia-

t ions.

5. Parameter values after each iteration.

6. Residuals of the projected points in their coordinate

system.

7. Variance-covariance matrix.

8. Standard errors.

AIt oui;h th is )r() t;ram is q,,ite accurate and p roduce: awealth ot inforna t ion, it i:; time consumin to utie. Thi:,coupled with the unavail:ll ility of calibrated ;,rid platesand steropilotter variety encouraged I,\(S to develop a

m i fi'ed ver ;ion. For in.;trument:! such as tht. Wild A-7,A-8, A-9 and A-10, the perspective centers do not vary withthe relitive orientat ion of the model -ince the :l i ofrotation f or o:ega and phi intersect the 1)rojection cardan.Tihe perspective centers change oniy if the instrument basecomponent is changed. Thus the coordinate determinationsneed be made only once per strip or, in special cases, perblock. However, for optical instruments such as the Zei:ssC-8 and Wild B-8, the project ion center varies with therelative orientation of each model since the axes of rota-tion for omega and phi do not intersect the projectioncardan. If the perspective center is to be determinedafter each model relative orientation, a quick method is

needed.

Instead of a reseau grid, the engraved lines and crosseson the plate carriers are measured and substituted for a

grid. These intersections can be quickly measured afterrelative orientation of each independent model. Theinstrument z and the numbering sequence must remain con-stant for each set of recordings. The program is designedto compute perspective centers of the left and rightcameras alternately. A maximum of ten points may be used,but six points are customary. The program output islimited to the x, y, and z coordinates of each projectioncenter, Dut this is sufficient input for the SIMBA adjust-ments. An example program execution is shown in Appendix B.

One's first experience with a large adjustment program canbe frustrating if two unique points have the same pointnumber. The block tends to fold in upon itself causinglarge residuals and an unsatisfactory solution. Anotherfrequent error is the exceeding of the bandwidth parameter.This is a result of long flight lines or incorrect modelsequencing in the program. SIMBA is unpredictable whenthis occurs and may diverge instead of reaching a solution.These and various other errors present in the data can bedifficult to locate. Also, it is expensive to run largeadjustments if errors are inherent in the data. For thesereasons the author wrote the SIMBA SORT program (SISRT).

It uses the exact same data deck (or file) as SIMBA and isquick and inexpensive to execute. The main program functionis to do a point sort and list the location by strip andmodel of each unique point. It also computes and lists thebandwidth of each point. This allows the user to insurethe SIMBA bandwidth Is not violated, or if it is, whichpoint is the culprit. A complete data deck/file listing

is provided. Sample program output may be found in Appen-dix C.

The Numerical Model Orientation Program (ORELM) was writtenby Randle Olsen of the U. S. Geological Survey in 1976.The purpose of the program is to use the omega and phi

elements from relative orientation, the desired model

scale, and the common omega, phi, airbase, and bz for

- i

ti ' ;' e:.: (EiK -' ) , :;.1, ' i Ii d - , ,K ra PC-2 ind th,.. ' -, '( )A ., iro v',t worki;i,,, to n c, rporit,

.h c , C n-3 nt i K i "r "l , pro i t ; ) p ,it i c r ;r :4 4,r-, 1li .i:, ,ii ; ,r a i x i. r7 v ,i-:ii q ,t e.. u t.

Ti. h.'. .I. l . ,I- o til. pr,'),r I .i i -I _' t ! cI . d t ,,ro , : ! ,: i i .' A n d I . ,. 1 1 ,. o r.. o t ' it 1 1 1. ); l I,: ;,,I r

.v i O U o :1d i: .\.Ap , :d i: D.

... y. ., -, 1 110'?h 7 i I.

It:4e rh!el :it Ion o th],'se st ' rioi and the the cooird

nares in Jction Space.

Aal:. ical and semi-canal tical methods are both capable ofproducing accurate results. Therefore, accuracy is not thejust .:a Tion for the extra expense of a comparator and then'-.naeri ;I ' prorams. However, there are inherent advantagesor*e -ttien o th te xposFirst of all, it ismheasior

nate in obc sce~.muhesa

Aa and -e3anayt : alds. One does not cave to wor

. r -I tn to remove oaralax from a model that is halfSecondly, a large variety of photography and cameras

can accommodated. Panoramic photography, for example,does not have a simple central projection and special cam-eras on high flying aircraft may have a very long focal

lenqzh which is impossible to duplicate on a stereo com-parator. Thirdly, analytical methods permit the input of

auxiliary sensors. Such sensors include stabilizationsystens, profile recorders, altimeters and astronomical

observations. This additional information may be incorpor-ated drectly into the solution and weighed according toreliability. It should be noted that some semi-analyticalprograms oow have this ability.

Analytical photogrammetry offers great potential. It canincorporate various distortions and deformations which areimpossible to duplicate in stereoplotters. The proponentsof analytical methods claim they should be able to obtainmore accurate results, but in practice this has not provento be true. The complex mathematics and equipment costshave frightened away many potential piactitioners. lAGS doesnot recommend one method over the other, but instead hasincorporated both methods into the software package.

When photogrammetric images are measured on a monoscopicor stereoscopic comparator, two dimensional coordinatesarc produced. The CRAST program refines these coordinatesand produces either GIANT input or independent model coordi-nates for input into SIMBA. Mathematical refinement of thecomparator measurements [s necessary because the model spacewas not physically reconstructed. It is mathematically

I

reconsr ricted , -;o we need to knew add it io na[ informat ionto accu rately dupIi cato what took place wh,_.n the photoswere taken. The first three items on the following listare required for program execition. The remaining itemswilI be used if they are known and input.

1. Camera Focal Length

- Approximate Fly Lig le igh t

i . Sohut Refraction Coefficient

Comparator Scale Corrections in x and y

5. Comparator Non-Orthogonality

6. Initial Airbase at Photoscale

7. Radial Lens Distortion Corrections

8. Principle Point Variation from Photo Center

9. -Calibrated Fiducial Coordinates

Depending on the information input and the desires of the

user, the CRAST program will correct the input coordinatesfor the following distortions:

i. Film Distortion

2. Earth Curvature

3. Light Refraction

4. Comparator Calibration Corrections

5. Radial Lens Distortion

If the GIANT option is used, CRAST output will consist ofa fit of fiducials to camera data, the refined plate coordi-nate listing and punched card output in the GIANT format oran optional format. If the SIMBA option is used, outputwill consist of independent model coordinates, y-parallax,model tie residuals and card output in the SIMBA or anoptional format.

The reason for the use of this particular program is itsadaptability to the lack of information which frequentlyis found to be the case in Latin America. For example,the program will accept the absence of calibrated fiducialsand even the absence of fiducial observations. CRAST wasoriginally written by Randle Olsen in 1973 and has beenrevised several times, including a revision by lAGS in1981. Sample program output can be found in Appendix E.

Th' Gene ra 1 Integ ratod An ai Lt i Cal Tr ian; iu at I on Pro,, ram

( 2][.\rNT) i:; desi-ned to pe frt orm a l,,ast :;q4.1r,2s idju:itm nt

Ot a block or 'rame photograph:;. It w:; written by Atef A.

r 1 a,, asa I in 1976 and hasa been adap t d for us;e by the I. S.

Geological Survey, the U. S, Forest Service and lAGS.

>anv variations of the program now exist and at Lhi';

writing, it is currently underg;oing yet another revision

at IAGS .

GIANT will anaLyt ical lv solve for the ground coordinatus of

image points measured on two or more photographs. It also

solves for each camera station position and orientation.

Only uncorrelated observations are acceptable and may be

individually weighed to reflect known precision. This system

allows the mixture .of horizontal and vertical control of

varying accuracies. It also allows the use of known camera

station parameters. The program propagates error estimates

through the solution, computes the a posteriori estimate of

the variance of unit weight and has an option for listingthe variance-covariance matrix and standard deviation of

the output parameters. These indicators are useful esti-

mates of the solution accuracy.

An initial approximation is required of each camera stationposition and orientation and of each unique image point.

The approximations of the image points are computed inter-

nally and gross approximations are acceptable for the camera

parameters.

Pro'ram dimensions require that no strip exceeds 20 photo-

grapIs or that there are fewer than ten strips in the

adjustment. IAGS currently has two versions of GIANT on

file with the difference being in the dimensions. The

principle version has the following parameters:

1. 460 photographs

2. 450 ground control points

3. 9509 unique ground points

4. Each point may appear on up to 10 photographs.

5. Object space is expressed in space rectangular

coordinates only.

The small version is similar with the dimensions reduced

to 150 photographs and 150 ground control points. The

large version requires 476K bytes of computer memory on theIBM 370 computer. The core capacity is reduced to 282K bytes

in the smaller version.

Experien~e has shown that several different types of execu-

tions prove helpful in data analyzation. The idea is

similar to that used in SIMBA. One must try to separate

the affect of different observations in order to locate

blunders. The recommended execution sequence is as

follows:

U the inter sectio-only opt ion nith no -r iund on-

r ,i . Ch er c r Iar ;e , o undt r-.

-the triangulation opt ion with ro round contrl,) I

oh -r or smaller error!3 in point ma rkig i n d ma;u:rin; ,es peC I Ill y in the tie oj) into.

I. U.o the triangulation opt ion with ;.'round control.

W c i1.ih t h image measurements tightly and allow the ground

control to move. Check for ground control errors and update

the exposure station parameters.

4. Use the triangulation option with full ground control.

Weigh the solution realistically; check all residuals.

5. This is the final execution. Use the triangulation

option, full ground control, realistic weights, and the

latest estimates of the exposure station parameters. Use

the error propagation option.

There are several things to watch for during the program

executions. Most must be learned through experience. One

important factor to remember is the bandwidth limitation.

It may be controlled by careful ordering of the frameposition and attitude cards. One can stack the photos

in any order. The trick is to keep the frames with common

Points as close together as possible. The pass and control

point selection procedures are the same as for SIMBA (see:L-ures 2 and 3). It should be noted that GIANT will notdistort block corners as severely as SIMBA, so corner posi-

tioning is not as critical.

The output listing includes:

1. Camera station parameters with residuals.

2. Plate coordinates with overall standard deviation of

x and y per frame.

3. Ground control coordinate listing with residuals.

4. Camera station corrections per iteration.

5. Triangulated residuals per frame for all ground

points.

6. Triangulated camera station position and orientation

residuals.

7. Triangulated ground point coordinates.

8. Applied ground control corrections.

9. Variance-covariance matrix and standard deviation for

the camera station parameters and ground control.

Excerpts from a typical program execution may be found in

Appendix F.

As with S I !',\A problems with -ii.;,puiiIhed or ': i ani:ibe rd dt apoiut s ind error,; in the or.. -an iLza t io n oC the data deck ,:ancause t iMC delays a-1d need Ic s: computer expe,,s e. A dataclean-up program was written by the author in 1980 to avoidsuch problems. This program (CISRT) is similar to the SIMBASort proram in that it uses the exaict :;ame input setup ;isGI.\NT and does a point by point analyzation. Th o outputcons ist 3 of the following:

1. Data deck listing.

2. Frame position and attitude card check.

3. A frame count by strip.

4. A point by point listing to include all locationsencountered anu bandwidth computation.

A sample of the program output nay be found in Appendix C.

SUMMARY

The programs presented here are not state-of-the-art tech-nology. They do'not have the most rigorous solutionsnor the most sophisticated computer manipulations. Whatthey do have is a good combination of flexibility, mathe-matical rigor and computer independence. The overallpacae can and will be improved. It is hoped that anew adjustment program written especially for mini-computerscan be incorporated into the system. The GIANT program isunder revision to make it more user oriented and to addadditional options. One such option would allow output ingeographics. The package is presented as a model for dis-cussion. Its greatest testimonial is that it is beingextensively used and has proven successful.

V.

LItass,iI Atet .. "General Integra ted Trian;'uIation Pro:gram,Unpub I ished User'z Guide, US Ceologic a1 S urvev-Topo-graphic Division, 1977.

6;hh, Sanvib K. , An l,.vt i clI 'hoto, ramme try New York:Pergamon Press, Inc., 1979, pp 1-5.F Hawk, James R. "Error Propagation and Least Squar s,"

Unpublished Instructional Material, 1978.

IAGS, ":)MA lAGS SIMBA, " Unpublished User's Guide, IAGS-TD,1978 .

lAGS, "Perspective Center Determination by Space Resection,"Unpublished Program Description, IAGS-TD, 1973.

Mikhail, Edward M., Unpublished Classroom Notes, Purdue

University, Department of Civil Engineering, 1978.

Olsen, Randle W., "Simultaneous Block Adjustment of Models,"

Unpublished User's Guide, US Geological Survey, 1979.

US Department of Agriculture, "Analytical Mapping System,"

Unpublished User's Guide, US Forest Service, Engineering

Staff, 1981-.

II

.4.1~

APPENDIX A

SAMPLE OUrPUT

FROMf SIMBA

I -~- - -- - -----------

iZi

I N

tti i i4

-~3 IDIIti 7 x 1 17 "I '7 ~ 7 i 7 1 ~

- p

IC T T T -

I T T -4

T %a N

a a ' n !

*1

I

TpII

4.

* I

-~

.4- I-

* - 4~ ,~ -J

'~ r- - ..

I IU

E I

__

I r r

I J, L I r t l1C; z I'

Al --, I . ;, l-

,Ci % c-C Irr~,~rN- 7'Cl Cj -4 t

3K rft4

f - 1 '1_j I j I1

I

Sj

1 c

r n c r 7 -4-

r

In

'k j r N o

-~z ~In I ~ f-c. '-ic r 'n

I ~ ~'d~*r 'Noc~ z oc

41 3L 1 L' !2 LN A~' S 'M N E N J

t 4x

K ..- -,-fi

K, z/.-~-.-T

rr

p rrI

'X Z .t -f C

r I'

3 1, 1 1

.C rf z

N, r n-

NJ:~ \c '

I:I

I T

III

j - C C

21 X Tl - I Nt~ r r r

A4 11

APPENDIX B

SAMPLE OUTPUT

FROM PROCEN

-t _ . 1'L. _________

4.y

,i

iIi

I LIi

APPENDIX C

SAMPLE OUTPUT

FROM SISRT

ti

Il

K 1 _1.

73 2

2 jr r -I

r 1- - ,r 7 . -.- T

*1 K 1

r j,

a;

•~ ~~ ~~ ".4 .° .- ." .. ... ..

d 4i ..... ..II

C r r14 : I I rrr ' r I

I

7 YTI I I I I

,- -w

-A

Ii I I.

-I I I

* i *

I IIp Ii

I I i

'I;4zr 1. ~ - I

~ 1 -\ 1 ~'

- 1"lI II I' II LII I

* .*4. ... ~. ~

i I

-. i,. I~rI~r-i9 -- q~ - S

~- Y -

\~ ~'J- -

i-jr'V -J

'~-~1~%~ ' -.2 - -I-

L.

I rIiI

ItI

- I -4'I .. AJ

J -

APPENDIX D

SAMPLE OUTPUT

FROM ORELM

... ..... ....--- ----

. . . 0 . )1 T

-~ - -~ -') -

LV 'I T IJ

c c

14 I

L 'i l c I clc

" I l

4* ' c C '-r .21-, r n P

'Z4J n I Z

'1co 141=4

47 a c cc cc zc ec

LII

T T

7-~ JD 7.

"I N ol 9c

71 - 'IC 3

m CIu- r 7 I MI

NJN

a -7 c 0': Y- J ,C

If Z~ J~~r~ C:. C 4

In P .Z 0C010 .d InI r r nr-

2 -r 1s * s 'lf 3

N. -j fn m In e7 L C.:: C c :cri= : t C2cY

k' In tr T c f-p

Ji1*lI 0 t~4 1 ~ .1 r~.T1 0

APPENDIX E

SAMPLE OUTPUT

FROM CRAST

I I I

" 11

i-I,

I I

I K

ii

I I

I I2

C I I

I I •i

I- J i II

i" I .... It .

1- I _ *I

I I

tA W

UJ

*0 I J

I 7>

Li cF II \

r '. . I l CI i4 -P X I

n In I

N r

r 7 Y- I, IJDr

di T .) Ii N d CX7* 1 . 1

F *~* ~ .'IT

I It

T, p II

f r

IJ, n

3x tlz

s 4 i

N Jz It I 1

II Il - 'I'

7 TI Ne

* :1=

3rI

t I I I

IN L i %

APPENDIX F

SAMPLE OUTPUT

FROM GIANT

r

~ I - ~ C - x-I cc

~ -~

lilt 11~111 Ii II II

fl

- I~u

I; if--~ ~ I-p

ii -~,-~.-- I.

:1 -i~-.~

2 -L~

Ill.

.- ~- ~t t-. I . *~ r ~lij ~.j

21

I..t 1*

K I

rx

~2

a a

Z

~ 4~ 1

4.21~

:~ -\ tJ~~~I

- I I

~,1p .4.

~*2

'-S -rp ri~i-

U, 1~' j::jII iiiiqt -

-J - F- F

I -P.- 1

- *1'~ I,' I F

I F

2-~ .~ I IIII F

1:

* I ~

~ -

~ K 1'- 4.

- -II- F.1 iI~4;: ~

IS.. - I III

gill -r

-J 7

I.

'71'-

;JI~4 1

It

~ '4I * -. 4., '1

'~ I '~ *~ * '4r 19 C -.4I I'-4

I r, -' ~J

z ~ ri

- I2'

Hr ~c 1 ~

a ~21 I2' Ir1AA I,

'a

4.'I ~ jA- .AJ

j ~.

~1*-A ~4Z~~>'

* '~ I!,':1

1 IllI C'

I ,~ Ii

~ ~(2 Al

aPm

'I

1~- I

'A- - -

A AL I- ~ ~.r T ~' ~ 4-

- I .. -

I I-'A - II-Ii,2

- 1 -]4. JiV'A T.1

*'A- - r~ 4'A r I I

'1. -~'IIt

'S C J ~ -4

.4--7

A-

~~-1

C

>1'p

A 2' rIi-,

I x'Jr *:~*-s Z

a -c

2

~ 12t

2

AV 1 ziF I ~ ]

* K -

K1 ii~

~f- II

'12 I U,,

-'-V

-, I

4 I I

Cl .~-2....~ I-

~ -~ \Jj'

I -lilt

4

I' - lJ.~

II '.1 '.1~

* I I - -~

.2V

- Jj.r

12. f I'C -

- '~- i-,. ,~

2 1. *T c

2 4

- -I 4.-c:a

*

-. *19 I t.

I,

', 4

*1 -~ -,

- C C %

I I

II

C I ja a. a. a! a. a

--

AJj

~z ~ F F I<A

'42~! I tI~

Ii. U IL III II II liii II~,I II II' II III II II -I.II

~I II

I j I

'1

11

E

4

I-

II

I .., I- - - °.

I-)

- Ii, _.:1

-'1

J

V,

L.

~1~d

I I

VL 'I

IA

x Jr.l

- 0

F 1

p. --

-~

-~ x~

2;

I.4.-' - -

I

22:2:

2

I S

t

'&

171

-Hpx r~--~- t1 Cr Ir4 -r,! AN .1 L :7 n

r z -7 -.1)

l, ,-.z- = - ._- r ...... ...

"r C -- - -.. . . ... . . - Z 7

4~- z

II

Z'rr sj =<, NJ -. - -

. . .C . . . . ." .r -

~~~--~~ -r- 5 VS-f flt f f C

do~~~ cs- o

z K' I4 pvp _

P I ~ 1 Nt .

~i~c 11 I c c., 00

c c C-I 'mD cn Cz,.n Z3'1

---- z z 7

1 1 7 Z

N I N2

SJ

I,

I I

-i [ I p m m~ l

J -J

4T

a I

I- I

-

a

- 4

APPENDIX G

SAM1PLE OUTPUT

FROM GISRT

Iii

I

Iii

III

-I I -

,=t _

_ . C

r :i

r 'r

r I I

I T 1I-- "

I4 1Irr ~ ~ ~ ~ ~ ~ ~ ~ - z *,- 4 J. 1 4

cI c 3-

- - "-1 : C- -t -C-C' -

I i I

... .. .. .. .. .. ... .. .. .. .. .. . ..... -.. -..-. '.. . . . .

I I

Ck C: C

. . . . . . . . . .I..

4: =

.- . . . . . . . .)~ .. . j p . . p * . * p* *\ p

't I1 C J, c= ~ j.-.

~ 4

r yP r' r . i- I., r-N, Z I . r? 4r r r t4 P ~ -1 rI r.Fl\' I-r4

J

r r L

j , t 40 7

Kj

...:; :-- - in r

" "

I.eij. j. .T'

T T -T i r T

-, $

Kl.i4~~i

-I

x ~ (~. - , -.

a.1

.1'

I I

£ rip 1-fl ~P t -, ~

QJ1£ - ~ 7

.4 -cv - - -. - X -. ~ - - -

C I r., c r c r~ p ri- -~ ~

I.: 7 ~ e 7Iz -

- -I ~ - '~~~'~-' C Pt ~s £' 4'~ 7 -, 7 7 C 7 ~ 3C '1---------------------

- r~. ~ - -

* I - - S T - - - - -

- - - 4 I.- C l~ -

- C - .~ ~ ~ ~ - C C C 7- - - ~ 1 C -. - -

±

- z - - - C - -

N

T K

IAT-

DI