Detection And Tracking For Time Sensitive Infrared Dim Small Target In Deep Space

With the development of modern military science and technology in the direction of space,the onboard infrared imaging detection system becomes more and more important,and the research on the detection and tracking of space-time sensitive infrared small and weak targets is also in urgent need.Compared with the common infrared dim small target,the space-time sensitive infrared dim small target has the characteristics of long imaging distance,fast moving speed,small and weak target,and the image has low signal-to-noise ratio,low contrast,and clutter interference.All of these have great challenges to the accurate detection and tracking of space-time sensitive infrared small and weak targets.In this thesis,the detection and tracking methods of infrared dim and small targets in space mission are studied.The main work and innovations are as follows:(1)In order to solve the problems of serious background interference and difficulty in detection of space-time sensitive infrared small and weak targets,this thesis studies the detection method of infrared small and weak targets based on human vision system,and proposes a multi-scale fractional entropy method,which introduces the fractional entropy into the detection of space-time sensitive infrared small and weak targets.First,each small window is calculated in the form of sliding window Then,the accurate position of the target is calculated by using the difference of the fractional entropy between the target area and the background area.The simulation results show that this method can solve the edge interference of the target in the cloud and the background,and get good detection results.(2)Aiming at the offset problem of kernel correlation filter(KCF)tracking method,a multi feature fusion KCF tracking method combined with high boost filter is proposed,which makes full use of high boost filter to suppress the background edge and makes the feature extraction of small and weak targets more accurate.At the same time,two features are used to train the filter respectively.Finally,the weighted fusion method is used to predict the final target location of the subject matter.The simulation results show that the method can reduce the offset of KCF tracking to a certain extent,and can still ensure real-time tracking.(3)Aiming at the classical spatiotemporal context learning visual target tracking(STC)method,this thesis proposes a tracking method based on the combination of double index edge preserving smoothing filter and STC.The double index edge preserving smoothing filter is used to suppress the noise and clutter in the spatial context of small and weak targets,so as to obtain a better spatial context model,which has a great effect on the above learning model in the whole spatiotemporal Promote.Through the test,simulation and evaluation of a large number of actual data,the algorithm has high tracking accuracy and stability.