IP LOGGED:

[Viewed 5054 times]

SITE URL 01: www.VoltsAmps.com
SITE URL 02: www.SVBxLABS.com
SITE URL 03: www.HERF.tk

"This site is powered by only passive components."

.:: Computer Vision Tracking ::.

Introduction:

Computer vision and tracking is becoming a hot topic these days as robotics and new cutting edge technology. Our computers are becoming more powerful and the users more demanding. Computer vision and tracking are now one of the military’s main points of focus as this technology is important in missiles, unmanned air craft and many other well known and even classified apparatus. In this experiment I will attempt to design a reliable object tracking computer system which will use my PC web cam as a video source.

Materials:

- PC
- Webcam
- Visual Basic or other compiler


Data:




The tracking software tracks my head as I move it around.



The software is quite flexible and has no trouble at all tracking multiple objects. In fact hundreds of objects can be tracked with very modest CPU useage.



In this animation the software tracks the wall clock as the camera itself moves about. The actual tracking is much higher frame rate than this animation.



This app allows the user to quickly and efficiently use their hand as a pointing device. A different algorithm that tracks specific outstanding colors can be used to monitor a user's hand position. The hand position can then dictate the location of the mouse cursor on the screen with no noticeable delays. The software even predicts the future location of the user's hand so that the scanning process can begin there and expand, to speed up processing.


Conclusion:

I was very impressed with my results after having tested my software's performance in object tracking. One limitting factor I noticed was the frame rate at which I took camera samples. My software was receiving about 2 to 5 frames per second from the camera. If the frame rate is increased to the maximum 30 frames per second supported by the webcam then very high speed tracking can be achieved, and tracking errors would be rare even at high object speeds.

© Copyright Rostislav Persion 2004