Abstract: In order to save energy of Wireless Video Sensor Networks (WVSNs), some compressive sensing methods are proposed to reduce the amounts of data for transmission. However, traditional compressive sensing (CS) uses constant measurement to sample images and can't obtain optimal sample data because of the sparse change of images. This paper presents a novel adaptive CS for multiple video nodes target tracking in WVSNs. In the proposed method, each active video sensor node uses an adaptive measurement and fusing method based on confidence weight to obtain optimal results. Experiments show that the method can effectively reduce about 85% amount of image data for fusion and achieve satisfy performance in real scenes.