Infrared Small Target Detection Algorithm Based On Two-Dimensional Matched Filtering

With the development of modern weapons, long-range strike has become an important tactical means in modern warfare. As a conventional equipment of remote combats, enemy aircraft, missiles pose a major threat to our important military equipment. If we can detect the enemy targets as early as possible, it can minimize the loss. In recent years, infrared imaging technology becomes a passive method of detection. It has all-weather, high hidden features, and it has been national attention and made a considerable development. Because that infrared sensor susceptible to atmospheric heat radiation and the role of distance and detector noise and other factors, targets in the infrared image appear as small low contrast targets, even as a point target, and often drowned in the noise. These factors increase the difficulty of detection and tracking of small targets. Therefore, carrying out in-depth study of infrared small target detection has important theoretical and practical value.Firstly, this paper introduces the background and significance of infrared small target study,and discussed its situation and development trend. Through the research and simulation of existing infrared small target detection technology, we point out the problems of the existing methods and the need for improved orientation.Secondly, the paper analysis gradient differences of small targets in infrared image with the surrounding background and diffraction characteristics of small targets. We propose an analog pulse curve meeting the spatial characteristics of small targets. Research and analyze the role of the two-dimensional Gabor filter template image in the image processing. And we give the methods and procedures to establish pulse image template.Then, the paper begins from the application of structure tensor in the image, and discusses the advantages of the structure tensor gradient pulse template. According to data characteristics resulting structure tensor, we propose an improved method of two-dimensional convolution. Experiments show that in the structure of tensor pulse image template extended after modification by the convolution method, gets the characteristics that odd-sized filter template having a center value of the maximum, center around information alternating positive and negative. It meets the gradient rotation invariant feature of small targets, and achieves the effect that target enhance targets and background suppression.Finally, this paper carries out the optimization work based on the proposed filter method.This paper through a large number of experiments and theoretical proof, obtains the optimal solution of the proposed filter parameters, and realizes background suppression of the infrared complex sky image. The results show that the two-dimensional matched filter based on a small target gradient background suppression algorithm has a significant role in improving the characteristics of small target SNR, SNR gain and relative contrast index.