Image processing has played a vital role in every aspect of human life. Video surveillance has reached a major out through by the application of advanced image processing and frame modeling techniques. Video surveillance is the most recent issue regarding community security and welfare. A monitoring system imparts the usefulness of imaging technology to detect, recognize and track a person or item of the suspect. Detection of behavior of moving objects in video streams is the a vital aspect. In this thesis, automatic real-time object detection and its behavior detection is implemented using Lucas Kanade and Harris Corner based approach. The velocity of various moving elements is reported in this article and its association with anomalous activity is also an inseparable part of this art. This work could be used to develop a surveillance system of static camera and robotic automation visual systems. Whenever a new object comes in the camera frame, a system uses the concepts of frame based processing using Lucas Kanade approach incorporated with Harris Corner Detector. The work presented here has been extended to work at video processing stage. In the later section, an investigation is carried out to define the optimum value of velocity parameters to define a behavior as anomalous. And it is analyzed that even though such an algorithm for video behavior detection at traffic points perform well but its accuracy can be further enhanced using three dimensional imaging and mapping of moving elements in a hyperspace.
Keywords—harris corner, spatio-temporal information, lucas kanade approach, anamolity detection.