![]() | Only 14 pages are availabe for public view |
Abstract In this dissertation, we provide an extensive review on human motion analysis and understanding. Then, we study and analyze the various effects of using different color space configurations with the mean shift algorithm on the tracking process (one of the stages involved in human motion analysis and understanding) in surveillance videos. We have used the gray space and four other color spaces. We have reached a conclusion that there is no color space configuration that can succeed all the time in all situations. Therefore, we provide a simple but rather effective algorithm for selecting the appropriate color space configuration based on maximizing the discrimination between the target object and its background. The proposed algorithm not only enables the mean shift tracker to succeed in situations where it already fails when persisting to keep a predetermined color space configuration with all situations, but it may also reduce the running time in some other cases. |