Compressed Sensing based background subtraction for object detection in WSN

Wireless Sensor Network(WSN) consists of tiny sensor nodes that are capable of sensing the environment and transmit the gathered information to the monitoring site. Conventional WSN architecture has limitations in bandwidth, power and storage. In case of multimedia applications data to be handled by the network is too large, which can be reduced with help of a technique known as Compressed Sensing(CS). In video surveillance applications, the main purpose is to detect the presence of an intruder or object in a particular area of interest. In traditional systems the entire video is transmitted in the network which consumes more energy, memory and bandwidth. These issues can be addressed with the help of a method known as background subtraction. In this paper CS based background subtraction is implemented for video surveillance which will reduce the storage, energy and bandwidth by transmitting the foreground measurements instead of all the samples. Performance of the framework is evaluated using parameters like precision, recall, F-score and accuracy. The proposed framework shows an accuracy of 99% indicating that the foreground object is detected correctly.