Security is becoming the primary concern of society due to a booming population, increasing scarcity of jobs, growing problems especially in urban cities, and the number of anti-social activities, etc. Having a security system is therefore becoming a requirement. Surveillance camera's output, when monitored, can track unauthorized objects from causing a menace. An important application of object tracking is video surveillance. In the proposed system, the object of interest is defined in each frame using a bounding box. The purpose is to determine the object's bounding box, automatically, in every frame that follows. It is a generic system for surveillance which indefinitely tracks an unknown bounded object, from online real time video or video file input, in an unconstrained environment. It works even in occlusions, illumination changes, and rotation and scale changes of object. Multiple packages offer tools, that can be used to implement computer vision systems, are available. One such package is EmguCV, which is used for developing this system. This is a C# wrapper for the OpenCV package which is in C++.

Long-term Tracking Learning from video Real-time.