Recognition and tracking of hovering object

The existing methods of monitoring may easily cause fatigue of monitor staff, just to obtain evidence for the most part. This paper provides a method of hovering behavior recognition and tracking, which can reduce workload and prevent something happen in advance. In order to realize the purpose, we utilize a video analysis technique including many automatic tracking algorithms, but these algorithms can not analyze the behavior of objects and have some defects. Therefore, we put forward the argument of hovering behavior recognition, and combine background subtraction with Camshift to realize tracing. The background subtraction detects object and approaches object contour with a rectangle which is also an object´s tracking box. Furthermore, Camshift algorithm is used to make up the shortage of the background subtraction method. The result of experiment shows that this method can meet the basic requirements of hovering tracking.