Compressed sensing based foreground detection vector for object detection in Wireless Visual Sensor Networks

Compressed sensing based background subtraction (CS-BS) plays a significant role in video surveillance applications in Wireless Visual Sensor Networks. This paper implements a CS-BS framework with a novel thresholding strategy to detect the anomaly with fewer measurements in a secured indoor environment. In CS-BS, the CS is performed on the difference frame which is sparse, thereby reducing energy, memory and bandwidth. In this framework, a foreground threshold is proposed based on the measurement matrix to extract the moving object from a scene. The performance of the CS-BS framework with FDV is evaluated using metrics such as detection accuracy, energy complexity, percentage of reduction in samples and measurements. The proposed CS-BS framework with hybrid matrix based FDV achieves around 95.8% reduction of measurements and 91% reduction of samples.