Research Of Low Slow Small Target Detection Technology Based On Infrared Search And Track System

Infrared search and tracking(IRST)system is a popular target detection and tracking device,because it has a passive working mode,and it can provide the early warning support in detection blind area of radar.So it is widely used as the complementary equipment of radar to detect the low slow and small target.The low slow small targets are always the targets that have great threat,so the performance of the low slow small targets detection is a very important index to evaluate IRST ability.Therefore,in this paper we propose a kind of target detection algorithm based on the practical working style of IRST.In the second chapter,the paper discusses the work of extracting the sky background,we propose an algorithm to distinguish the all-sky background image,the all-ground image and mixed background image.By using this algorithm we can pick out all-sky image and mixed image.Then we studied the method of extracting the sky part from the mixed background image.Because the effect of segmentation by traditional threshold segmentation method,or edge detection and region growing algorithm based on IRST captured image are not very good.This paper proposed an algorithm to pick out sky part of infrared image by local description using Gabor wavelets.Through experiments we can find that the method has good reliability and strong practicality.In the third chapter,we discussed the single frame target detection algorithm,we propose a kind of target detection algorithm based on small target model.The algorithm can detect the real target and the number of false alarms introduced under low threshold is not too much,and it has some inhibition effect on the disturbed area targets and the remanent ground part of the infrared image.And the target of the algorithm can keep the imaging features of the small targets,so it will form solitary points distribution in the image,which provides support for the multi frame detection.In the fourth chapter,the article studied the method of finding real targets under the condition of many potential targets.According to features of higher frame rate of the IRST under normal circumstances,the real targets in multi frame images show uniform motion style of straight line,whereas the noise in the multi frame image presented random distribution.According to this difference,this paper introduced the random set model to describe the potential targets,and proposed the multi-step PPU filter algorithm and the multi-step LPPU filter algorithm to monitor the potential targets and extract the real targets which having uniform motion style of a straight line.