Header menu link for other important links
A Hybrid Vision System for Dynamic Obstacle Detection
Shankar Sharma Kirti, Ranjan Sahoo Sudeepta,
Published in Elsevier BV
Volume: 133
Pages: 153 - 160

This paper deals with the depth analysis of a target in the dynamic state using combined techniques of stereo vision and Kanade-Lucas-Tomasi (KLT) feature tracking based method. The stereo vision system built using low-cost cameras helps in finding the depth of points on target; however, it is incapable of capturing all the depth point details on the target. Similarly, Kanade-Lucas-Tomasi (KLT) feature tracking based method provides only the direction of displacement of the target without quantifying it. Hence, we propose to develop an algorithm which fuses the techniques of stereo vision method and Kanade-Lucas-Tomasi (KLT) feature tracker to track the dynamic target with static observation point. MATLAB based point cloud generation is used frame-by-frame to map the three-dimensional environment by using the self-integrated low-cost stereo vision system. An initial frame is used to define the target to be tracked. The proposed algorithm is tested on vehicle movement with various speeds and directions. The algorithm is validated and verified for its performance and accuracy by comparing the experimental outcomes to the actual displacement of P3-DX. Results show that the experimental results are within the acceptable tolerance.

About the journal
JournalData powered by TypesetProcedia Computer Science
PublisherData powered by TypesetElsevier BV
Open AccessNo