A New Method to Detect Moving Targets

Abstract

In the moving target detection, there are a lot of methods. But the effects of those algorithms are limited, especially in complex environment. This paper putts forward a new method to detect moving targets, which uses virtual binocular disparity. It is called virtual because the environment uses one camera but the theory is about two cameras. Virtual binocular disparity algorithm regards moving targets as static targets under two cameras. It uses two neighboring images, the current frame image and the next frame image. The algorithm looks the two images as two static images under two cameras, then uses stereo match theory to calculate the disparity of the neighboring images. Because the background of the two images is the same, the disparity of their background is about zero. And the disparity of moving targets is positive or negative. The sign of disparity means the direction of targets moving. So we can use the sign of disparity to distinguish the targets moving in different directions. The direction detected out will be helpful for subsequent work, for example, targets tracking and counting etc. This algorithm fully uses the continuity property of moving targets, so it can have good effect. In the end of the paper, experiments are used to prove the algorithm.