| Video based detection systems rely on the ability to detect moving objects in video streams. Video based detection systems have applications in many fields like, intelligent transportation, automated surveillance etc. There are many approaches adopted for video based detection. Evaluation and selecting a suitable approach for pedestrian and vehicle detection is a challenging task. While evaluating the object detection algorithms, many factors should be considered in order to cope with unconstrained environments, non stationary background, different object motion patterns and the variation in types of object being detected.;In this thesis, we implement and evaluate different video based detection algorithms used for pedestrian and vehicle detection. Video based pedestrian and vehicle detection involves object detection through background foreground segmentation and object tracking. For background foreground segmentation, frame differencing, background averaging, mixture of Gaussians and codebook methods were implemented. For object tracking, Mean-Shift tracking and Lucas Kanade optical flow tracking algorithms were implemented.;The performance of each of these algorithms is evaluated by a comparative study; based on their performance such as ability to get good detection and tracking, CodeBook algorithm is selected as a candidate algorithm for background foreground segmentation and Mean-Shift tracking is used to track the detected objects for pedestrian and vehicle detection. |
