Moving Object Detection Under Complex Background

Line array CCD vertical target measurement system can realize the test of the coordinate position of the low ballistic target. Before the accurate calculation of the projectile targets' space position, the image processing algorithm is used to accurately detect the projectile target from the line array CCD image acquired. However, the range of outdoor environment is complex and changeable. In the image acquisition process, there will be a lot of interference factors, such as noisy outdoor environments, light intensity of illumination changes, shaking trees, mosquitoes and so on, if the general target detection algorithm is used to detect target, these interference factors will lead to failure, and ultimately can not locate the target coordinates. Therefore, in the process of application, it is necessary to improve the moving object detection algorithm based on a variety of special problems, and increase its scope of application.In this paper, a new algorithm for moving object detection in complex background is studied. The main task is how to minimize background interference when moving objects in the image under complex background and special environment, the final accurate target detection.This paper mainly based on linear array CCD image in complex background dynamic small and dim target detection research, incorporate the attention mechanism of human visual system to the dynamic small targets detection, and puts forward two kinds of target detection algorithm to solve the special problems that the moving target in the image is submerged in the dynamic complex background, and it is coincident with the swaying leaves in the background which is difficult to detect by the common method. The first algorithm is a moving object detection algorithm based on the difference of visual saliency, and the other is an improved algorithm for moving objects detection based on visual saliency. The core idea of the first algorithm is that according to the characteristics of line-scan CCD target image, all static and most slow movement complex background is removed through computing the difference of two near frames in the image, a new feature vector is constructed, double search windows are designed which is similar to the target shape, and the line-scan CCD image's saliency map is got through calculating the similarity of feature vector of the center and surrounding pixels'neighborhoods in the window, then the saliency map is clustered and segmented to extract the targets. In order to reduce the complexity of the algorithm, the improved moving object detection algorithm based on visual saliency is put forward. In the algorithm, it will no longer deal with the image difference, instead of using linear CCD image features, fusing the difference idea to the computation of the saliency map directly, that is, because there is a correlation between the adjacent rows, changing the double window size from M1*N1—M2*N2 to 1*N—M*N, only calculating feature vector similarity between rows to express each pixel to a significant degree, and get the image's saliency map, and according to the shape feature of the improved window and the feature of the image, a new feature is proposed to build a new feature vector.Based on the above algorithm, the validity of the algorithm is verified by experiments, and compared with the common methods. The results show that the moving object detection in static background of different light intensity and in complex dynamic background, this algorithm can suppress the noise and background, detect the target, and achieve the desired results.