The Research And Imple Mentation Of Algorithms Of Infrared Small Targets Detection Under Sky Background

Infrared Small Target Detection technology can be good make up for the defects ofradar detectors; therefore there are many researchers research it, this paper also studiesthe Infrared Small Target Detection technology in sky background, the main tasks andresearch are as followings:(1) Analysis the IR target and background model, and studies the difficulties of thekey technology of the current IR dim target detection in-depth in sky background.(2) Research space and frequency domain background suppression techniques,and background modeling techniques, analyze the scope of each algorithm, andremoved the intermediate values of the maximum mean template based on themaximum mean filtering, got better effect of the improved filtering algorithm, and thedim target were more prominent in the residual image; because the traditional filterswere isotropic, they were not well adapted to the edges of the image, leaded to the edgesof filtered image were blurred, so this paper introduced the anisotropic diffusion filterto build background modeling of the infrared image, because it could adaptivelyadjusting filter coefficients, so got a good background modeling, as well as a goodtarget detection results. Because the threshold value and variance of the traditionalSusan filter need manually set, and once set can not be changed, but the variance andthreshold completely different in all regions of image, so the filtering effect was notideal, this paper summarized the previous work and proposed an adaptive Susanfiltering algorithm, through using the mean and variance of the local area with imageinstead of the threshold and variance parameters in original algorithm, it could avoidmanually set, and could be a good reaction parameters of the different nuclear regions,so it obtained better effect than the traditional methods of background modeling,Introduced the image signal-to-noise ratio, the image improved signal-to-noise ratio,image signal-to-clutter ratio, the image improved signal-to-clutter ratio, background suppression factor to test the various algorithms, found that the improve Susan filteringoperator had the best performance.(3) Used the fixed threshold, adaptive threshold and the OTSU thresholdsegmentation algorithm to split the detection results, segmentation results can be seenthe three algorithms split results contain a large number o f false targets, they affectedthe dim target detection seriously, therefore, this paper present the secondary judgmentby threshold segmentation, first, use the three segmentation algorithms to split thetargets, then use the second judgment on multiple split targets, only meet the certainconditions, the target is the infrared dim target, experiments show that it can greatlyreduce the false alarm rate.(4) Analysis the Hough projection of linear motion detection, because the Houghprojection points were too much to affect the accuracy and efficiency of the Houghprojection, so this paper present the Hough projection based on region and least squaresfitline detection algorithm, the first algorithmanalyzing the area projection point fromalarge number of projection point, thereby improving the accuracy and efficiency of theprojector via reducing the Hough projector number of points; another algorithm was tofit the large number projection points, to avoid Hough projection, it also can greatlyimproved the efficiency and accuracy of the algorithm.(5) The traditional dynamic programming algorithm was generally used forinfrared dim target detection tracking, by summing the multi-frame infrared imagepixel values, and setting the threshold, thus get the segmented motion infrared dimtarget in consecutive frames, because the gray values of the background of the infrareddim target may be very high, the accumulation process may not be able to detect dimtarget, therefore, this paper built the background modeling of the infrared backgroundfirstly, got rid of the high brightness of the background, and then used the dynamicprogramming in the multi-frame residual images, cumulative multi-frame pixel valuesin the residual images, which ultimately detected the motion infrared dim target.Introduced the tracking before detection algorithm to the high mobility of infrared dimtarget, it can capture the trajectory of the infrared small target, thus verify thepracticality of the algorithm.(6) According to the content of this paper, we designed and built the infrared smalltarget detection in sky background software algorithm simulation platform, anddescribed the functions and use of the various software modules in detail, facilitated thesimulation and results display of various algorithms.