The Research Of Weak Target Detection Under Complex Background

In weak illumination scences, the image catched by image senors has complex backgroud, which mainly consists of gray level information, loses chrominace information, and the texture imformation is also not obvious. In addition, long distance moving targets are usually characterized by time-domain non-stationary signal in two-dimensional imaging plane, so the methods of detecting dim target in complicated background should take advantage of the image spatial information rather than time-domain information. Edge is the most fundamental and abundant information in image spatial domain. Consequently, the target can be separated from background by extracting the edge so as to realize the detection of dim target.The paper studys the technologies of dim target detection in complicated background, the main content of this paper is summariized as follows:1. Edge detection is introduced because of distance target’s non-stationary feature in time-domain. Firstly, the Canny operator mainly introduced in this paper is improved based on the analysis of the first and second order differential operator. Whereafter, the morphological and wavelet edge detection are analyzed and implemented in order to contrast performance of before and after improvement.2. Generally, edge tracking can get smoother curve of edge information. In this paper, the dissimilarity measure on the selection of edge tracking direction is improved, adding pixel gray level and gradient information to the measure makes correlation between adjacent edge pixel closer, which can track to smoother, more accurate path of the edge and improve the effctiveness of edge tracking. Besides, against the end condition of edge tracking, a new method of breakpoint judgment is adopted. The experimental results show that the mothod can effectively handle isolation and burr edge points.3. The Hough transform and its detection method of circle are analyzed. For edge extraction has great influence on the result of detecting curve, an edge tracking method of dissimilarity measure based on the idea of Canny operator is proposed to connect edge section, and then Hough transform is used to detect curve on the connected edge. The experimental results show that this method can effectively restrain the influence of noise and clutter background, as a result, the false alarm point of target detection is reduced.4. Finally, the improved algorithm is implemented and compared with the former algorithm. The video image acquisition system is vulnerable to clutter background, camera shake and transmission line. Therefore, several image preprocessing methods are analyzed and implemented. To illustrate the availability of edge spatial information, time-domain detection methods analyzed and implemented in this paper are compared with the improved algorithm. The experimental results show that the image’s edge information should be utilized rather than time-domain information when the video image has low signal noise ratio, the target’s time-domain motion characteristics have no rules to follow and the gray transition of target’s edge is strong.