Background Suppression Based Small Infrared Target Detection

With the rapid development of computer and digital image processing technology, infrared target detection technique has been used in many applications, such as astronomy prognosticates, remote sensing, target tracking, guiding etc. Infrared small target has the characteristics of far distance, small area, weakness of shape of image, lose of detail of image and low SCR etc, the research of infrared target detection of theories and methods is concentrated on rising of SCR, the detection ability of algorithm,coverage,efficiency of computation and reliability of detection algorithm. In this paper, the detection of small infrared targets under the sky background is discussed, and the main research contents of this thesis are as follows:Firstly the paper studies the image preprocessing problem of infrared small target under complex background. It discusses the suppression effect of traditional image preprocessing method on the infared image background, aiming at finding a new train of thought on the basis of the existing pretreatment methods, which can describe and process different background, to achieve a better background suppression effect.Then according to the characteristics of gray level distribution in different areas of infrared images, local variance is used to describe the image dispersion. The paper puts forward a kind of processing of idea using double local variance to deal with infared image. After analyzing the feasibility of this way, different regions of the image are classified according to their characteristics, establish classification model, the optimal template in the method is obtained by quantitative calculation of the local variance of the classification model, and then finish the image preprocessing. By comparison, this method not only suits for different kinds of sky background but can greatly improve the image of SCR, and the effect on complex background suppression is very obvious.Finally, image segmentation is studied. First of all, several thresholding methods that are frequently used are introduced. Unfortunately, with none of these methods can we get an ideal result when preprocessing infrared small target image. Because one of the obvious features of infrared images is uncertain and unclear, the image is fuzzified, and the target detection in the image turns into a question of fuzzy processing, considering the ultimate goal we acquire by the human eye to see the information as a benchmark, a new image segmentation algorithm which combines one-dimension maximum fuzzy entropy and HSV(Human Visual System) is put forward, the gray information of pixels in the image is transformed into the luminance information of the human eye, so we can acquire the information of the target from the angle of HSV. By contrast, the algorithm the article puts forward has a higher detection rate and it can meet the requirement of image real-time detection.