Moving point target detection based on spatiotemporal image

In view of the limitation existing in detecting moving point targets in complex background, a new detection method with detailed theory based on spatiotemporal image is presented. In 2D spatiotemporal image, moving object shows finite line segment and complicated texture of background displays infinite straight line. In accordance with this characteristics, moving objects can be effectively extracted with local auto-adaptive threshold, and the trajectory of moving object can be gotten by applying projective transformation from spatiotemporal space to geography space. Experimental results show that the method can effectively detect low SNR point targets in complex background at the speed of less than 1pixel/frame. Furthermore, multi-target detection can be performed simultaneously.