Intelligent Compression Method And Real-time Implementation For Infrared Remote Sensing Reconnaissance Dim Target Image

With the development of space exploration and earth observation in China,image resolution and frame rate are getting higher and higher,which leads to the contradiction between the amount of data transmitted from satellite to earth and the limited down transmission bandwidth.The improvement of image compression performance becomes the key to satellite data transmission.The research content of this paper involves the design and implementation of the infrared remote sensing reconnaissance image data compression system.In view of the contradiction between maintaining the target information and improving the compression ratio,as well as the problem of designing high real-time system under the condition of limited resources on the satellite,in this paper,by analyzing the characteristics of the image with targets,an Object-Oriented Intelligent compression method is proposed,which consists of two steps: weak and small targets detection and block compression.The algorithm research and improvement are carried out respectively,and a high real-time intelligent compression system based on FPGA is designed.For the detection of small and weak targets,this paper is based on the multi-scale block contrast measurement algorithm(Multiscale Patch-based Contrast Measure,MPCM),this paper compares the characteristics of traditional infrared small and weak targets in detail with the characteristics of scanning image of small and weak targets researched in this paper,In view of the influence of stripe noise in scanned image and multi-scale and non-uniformity of image in large scene on detection,two improved schemes,stripe noise suppression and multi-scale multi threshold segmentation,are proposed and integrated into an infrared weak and small target detection method IMPCM(Improved Multiscale Patch-based Contrast Measure)for large scene scanned image.After verification,for the image with stripe noise,the performance of IMPCM in SNR gain and background suppression ratio is improved by about 2-3 times;for the image with complex background,the false alarm and missed alarm are greatly reduced,and the accuracy and recall rate are significantly improved.To solve the problem of image compression,JPEG-LS compression algorithm is used.Based on the requirement of error resilience,block compression is needed.In this paper,the reason of the decrease of compression performance is analyzed,and then a block compression method using local low dynamic range to adaptively adjust BPP compression parameters is proposed,which can effectively improve the compression ratio of remote sensing image.Furthermore,by combining IMPCM target detection algorithm and block compression method,the framework of intelligent compression algorithm is built.Compared with JPEG-LS block lossless compression method,this method can increase compression ratio by 3 to 5 times while keeping target information.Finally,aiming at the low resource consumption and high real-time requirement of satellite system,the FPGA system architecture and function modules of intelligent compression algorithm are designed and optimized,and the board level development and ground test platform of intelligent compression system are built.It is verified that there is no missed alarm in the typical data test in this scenario,and the average compression ratio is 5.7167,the compression pixel rate is as high as 38.36 mpixel / s,and the compression delay less than 1ms.The system level verification is completed to meet the needs of the project,and the effectiveness of the results of this paper is verified.