Pattern Matching For Tracking Applications

A New method is developed to improve an existing recognition system for video streams by using knowledge of object features occur and do not occur in subsequent frames was used to  filter false objects and to better identify real ones. The recognition ability was tested by measuring how many objects were found and how many of them were correctly identified in two short video files. The tests also looked at the number of object detections. We have developed a system which empowers users to upload MPEG video files and track the object using the web based front end. Evaluations are accomplished by comparing tracking output objects with a set of ground truth objects according to different performance metrics. These metrics are object based as well as speed based. Profiling has been done on the program to evaluate the gain in hardware realization of the algorithm.