Khandaker Abir RahmanKiran S. Balagani

Vir V. PhohaChuka Okoye

15 June 2009

Objectives

Related Works

Contributions

Proposed System

Experiments and Results

Conclusions

Future Directions

To locate eye position using CCDcamera.

To calculate human interocular distance(the distance between eyes).

To measure person-to-camera distance.

To measure movement speed in real-time.

Distance Measurement

Contact and non-contact approaches.

Image-based measuring approaches.

Speed and Motion Detection

Methods for detecting human motion includebackground subtraction [1], templatematching [2], optical flow [3] and temporaldifferencing [4].

This system measures the person distanceby simply analyzing the interocular distancewithout using extra CCD cameras, flashlights, laser pointers, sound reflection toolsetc.

This system also measures the movementspeed of a person based on the interoculardistance.

Proposed System

Image Pyramid

Generator

Assessment

Function

Calculate

Original Face

Position

Output face

with eyesInput face

Face

Pyramid

Generated eyes

Eye Mask

Locate the pixel

coordinates of the

eyes

Find the Euclidian

Distance between two

pixel coordinates (in

pixels)

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

0 10 20 30 40

Object to Camera Distance (in inches)

Sq

uare

of

Eye D

ista

nce (

in

pix

els

)

0

2000

4000

6000

8000

10000

12000

14000

0 10 20 30 40

Object to Camera Distance (in inches)

Sq

uare

of

Eye D

ista

nce (

in

pix

els

)

0

2000

4000

6000

8000

10000

12000

0 10 20 30

Object to Camera Distance (in inches)

Sq

uare

of

Eye D

ista

nce (

in

pix

els

)

0

1000

2000

3000

4000

5000

6000

7000

8000

0 10 20 30

Object to Camera Distance (in inches)S

qu

are

of

Eye D

ista

nce (

in

pix

els

)

(a) Height over 5'8'' (b) Height ranging between 5' 4'' and 5' 7'

(c) Height ranging between

5' and 5' 3''(d) Height ranging bellow 5'

edMAX

0

2000

4000

6000

8000

10000

12000

0 10 20 30 40

Camera Distance (in inches)

Eye D

ista

nce

2 (

in p

ixels

)

(y coordinate)

edMIN (x coordinate)

GMid (x coordinate)

DB

Look for height

information for a

specific person

Height found?

Default Camera

Distance-Eye

Distance

Function0

5000

10000

15000

20000

0 10 20 30 40

Camera Distance (in inches)

Eye D

ista

nce (

in p

ixels

)

Select function

based on

height

Yes

No

5'8'’<=h<=6'

0

5000

10000

15000

20000

0 10 20 30 40

Camera Distance (in inches)

Eye

Dis

tan

ce (

in p

ixel

s)

4'8'’<=h<=5'

5'3'’<=h<=5'7'’

………

Camera Distance

The movement speed can be measured by the following equation,

where is the speed at , at .

1

1

ii

ii

itt

dds

is i

t '1 ci

dd it

Experiments & Results

0

20

40

60

80

100

120

0 10 20 30 40

Actual Camera Distance (in inches)

Accu

racy (

%)

0

20

40

60

80

100

120

0 10 20 30

Actual Camera Distance (in inches)

Accu

racy (

%)

0

20

40

60

80

100

120

0 10 20 30

Actual Camera Distance (in inches)

Accu

racy (

%)

0

20

40

60

80

100

120

0 10 20 30

Actual Camera Distance (in inches)

Accu

racy (

%)

(a) Height over 5'8'' (b) Height ranging between 5' 4'' and 5' 7'

(c) Height ranging between

5' and 5' 3''

(d) Height ranging bellow 5'

In this system, a simple image-based personto camera distance and human movementspeed measuring system is proposed.

The proposed system is simple, costeffective and efficient for real-timeimplementation for human motion detection.

The overall accuracy of the person-to-camera distance measurement system is94.11%.

Human motion detection for video surveillance.

Accuracy measurement of the human motiondetection and movement speed measurement.

Consideration of side face views and facerotation for improving the accuracy ofmeasurement.

Human height and weight which influences theinterocular distance needs to be addressed.

3-Dimensional interocular distance considerationfor more robustness.

[1] S. McKenna, S. Jabri, and Z. Duric, "Trackinggroups of people, Computer Vision and ImageUnderstanding," vol. 80, pp. 42-56, 2000.

[2] A. Lipton, H. Fujiyoshi, and R. Patil, "Movingtarget classification and tracking from real-timevideo," In Proc. IEEE Workshop on Application ofComputer Vision, 1998.

[3] D. M. Gavrila, "The visual analysis of humanmovement: A survey, Computer Vision and ImageUnderstanding," vol. 73, pp. 82-98, 1999.

[4] B. Jung and S. Sukhatme, "Detecting MovingObjects using a Single Camera on a Mobile Robot inan Outdoor Environment," The 8th Conference onIntelligent Autonomous Systems, pp. 980-987,2004.

Thank You