Study On Methods For Moving Point Target Detection In Image Sequences

Detecting and recognizing the dim moving point target from image sequences in low SNR environments is always a key technique for moving target detection and tracking, it is extremely significant to remote supervisory and control system of both sky and ground. This paper focuses on image preprocessing and target multi-frame detection in this research area.At the beginning, several general image preprocessing algorithms is recommended, meanwhile, the background suppressing method based on mathematical morphology filtering is discussed in detail and deeply. Moreover, though processing the real images by this morphology filtering and other methods, it shows us that morphology filtering has a lot of good qualities, which are excellent performance and fast operation.Then after studying of representative target detection methods, a new algorithm based on directional cumulation is proposed to detect the moving point target from image sequences in this paper. First, single frame detection is performed, and then the logical operation of multi-frame images are used to obtain a binary planar projective image. With that based on the continuity and discipline of the moving trajectory of the point target, directional cumulation is used to carry on the second detection of the target and the parameters of its moving trajectory can be acquired by Hough transformation in the last, The result of experiments shows that the proposed method is effective to detect and track targets moving along linear or curvilinear trajectory under the condition of unrestricted velocity in low SNR, and it is of low complexity.