Design And Implementation Of Multimodal Pre-detection And Compression System For Space Based Infrared Dim Small Target Image

With more and more spacecrafts entering space,space monitoring and situational awareness have received wide attention.Space-based infrared surveillance is an important means of space environment and target surveillance.Through the satellite-borne infrared imaging system,important space targets(such as orbiting spacecraft,space debris,space stars,etc.)are accurately detected,tracked and monitored,space target characteristics are collected,and space targets are cataloged and managed.Among them,strong real-time and reliable satellite-to-ground transmission of large-scale,high rate and high dynamic space-based infrared surveillance image data is the key technology to realize the task of space target surveillance.In this thesis,the multimode compression method and system implementation technology for small and dim infrared target image on satellite are studied to achieve real-time compression and transmission of infrared surveillance images with limited download bandwidth.For infrared small and dim target detection,the characteristics of large-format space-based infrared images are analyzed.By comparing and analyzing the advantages and disadvantages of typical detection algorithms,MPCM(Multiscale Patch-based Contrast Measure)algorithm is selected as the basis to study its limitations in large-format infrared images.Aiming at the two problems of increasing false alarm caused by cloud edge and decreasing detection rate caused by large format highlight background,AMPCM algorithm is proposed.The experimental results show that AMPCM has a higher detection rate than MPCM algorithm,and the signal-to-noise ratio gain and background suppression factor are increased by about 4.2 times,and the false alarm rate is reduced by 1/7 of MPCM.For Space-Based Surveillance image compression,based on the LOCO-I(Low Complexity Lossless Compression for Images)algorithm,aiming at the problem of low bandwidth constraints,this thesis uses the Region of Interest(ROI)compression method to divide the image into the area of interest and the background,which can significantly reduce the data volume by reducing the size of interest region and the frame frequency of the background data.In combination with detection technology,a pre-detection + ROI compression method is proposed,which uses AMPCM to pre-detect the data in the area of interest and background region,lossless compression to ensure information integrity in the target area,and near-lossless compression to reduce data volume in the non-target area.However,there are two problems of large cache requirement and poor single threshold performance in background data detection process.A pre-detection + ROI compression method for background partition detection is proposed to solve the problems.The experimental results show that the improved compression method improves the compression ratio by about 4 times compared with the global lossless compression and effectively alleviates the bandwidth pressure.In the aspect of system implementation,FPGA is used as the platform to design multi-channel parallel high-throughput architecture and complex data flow scheduling method under the limited resource constraints,realize multi-mode pre-detection compression system,and build a ground test platform for verification.The results show that the system delay is less than 7 ms,and the data throughput can reach 115 Mpixel/s,which meets the requirements of tasks.