Infrared Small Targets Detection And Tracking In A Sky Background

With the development of infrared detector technology,infrared detection,as an effective complement or substitute for visible detection,has been applied more and more widely in various fields.Target detection and tracking technology based on target infrared characteristics has also become a hot research topic.Usually,the infrared target has better intensity and contrast than the background,but in the practical application of the infrared target detection and recognition system,If the target is from far and near,it is always hope that the target can be captured from a long distance and the real target can be locked;if the target is from near and far,it is hoped that the target will not be disturbed by the cloud or be occluded by the cloud,and can be captured again after cloud emergence,thus extending the detectable distance of the stable tracking target.Therefore,the target based on the long-range limit becomes the characteristics of small and weak targets,and its effective detection and stable tracking become the key technology to improve the performance of equipment.Based on the above requirements,this paper mainly focuses on two key technologies of infrared dim small target detection and target data association under cloud interference.The main work is as follows.(1)The characteristics of infrared dim and small target image are analyzed,and the main difficulties of infrared dim and small target detection and tracking under sky background are summarized.The pretreatment technology of infrared image is studied,and the median filtering algorithm with template size of 1*3 is used to process the isolated noise points in infrared image.Top-hat algorithm,multi-structured element Top-hat algorithm,spectral residual algorithm and multi-scale gray difference algorithm are used to suppress the background of infrared image.By comparing the processing results,multi-structured element Top-hat algorithm is selected to suppress the cloud background.(2)In order to improve the correct detection rate in low-contrast infrared images,an infrared dim small target detection algorithm based on multi-scale space and trajectory correlation is adopted in the infrared target detection under the actual cloud background.The algorithm uses Gauss Laplace operator to build multi-scale space,searches local extremum points to get the position and size of candidate targets,and then uses inter-frame correlation information to eliminate false targets and confirm the real targets.Experiments show that the method is effective in correct detection rate and false alarm rate.Finally,the hardware logic of the detection algorithm is designed to realize the transplantation of the existing platform on the FPGA.(3)In order to improve the tracking accuracy and reduce the ID conversion rate in the infrared multi-target tracking process,this paper uses the infrared weak multi-target tracking algorithm based on Kalman filter.The algorithm uses the threshold segmentation method to obtain the location and other information of the target.Two classical yet extremely efficient methods,Kalman filter and Hungarian method,are employed to handle the motion prediction and data association components of the tracking problem respectively.The superiority of the algorithm in tracking accuracy,ID conversion rate and speed is proved by experiments.