In automated video analysis, the performance of the detection stage is integral to the performance of subsequent stages such as tracking and behavior recognition. In this paper, we propose a method to boost the speed and accuracy of Wu and Nevatia's edgelet based detectors, by incorporating motion information to isolate active regions before actually performing detection. Experimental results show that the detection is significantly improved by using our method.
Published in:
Computer Applications and Industrial Electronics (ICCAIE), 2010 International Conference on
Date of Conference: 5-8 Dec. 2010