Improved Frame Difference Algorithm Based on CNN for Moving Target Detection

Abstract

In order to improve the shortcomings of frame difference method in moving target detection and improve the accuracy and robustness of moving target detection, an improved frame difference method based on CNN is proposed. Besides, in order to solve the problem of blind spot in autopilot and improve the safety of autopilot, SIFT algorithm is used to obtain the panoramic image of moving target. Firstly, frame difference method is adopted to obtain the foreground image, and then CNN is used to repair the foreground image. Finally, SIFT is used to splicing the restored foreground image to obtain the panoramic image of the foreground target. Experimental simulation results show that the proposed algorithm is effective, efficient and accurate in the extraction and repair of moving targets. With the proposed method, a complete panoramic image of moving targets can be obtained.