Automatic detection of entry into a restricted area

Detecting people and objects entering and exiting a region of interest is an important problem in the area of video surveillance and is typically called virtual fence detection or trip-zone detection. This work proposes two methods to address this problem. The former method deals with cases where a person or object is crossing a horizontal virtual fence while the person or object is moving in a horizontal direction whereas the latter deals with crossing of virtual fence while the person or object is moving in vertical direction. In the first method, crossing of fence by objects as well as persons is detected through simple temporal differencing based motion detection. To detect objects climbing up or down a fence, a spatio-temporal method is proposed. The proposed methods are evaluated under various real life scenarios and results demonstrate the efficiency of the proposed model as well as the real time performance.