FPGA Design And Implementation Of Remote Sensing Cloud Image Compression Algorithm

With the rapid development of remote sensing space industry,the limited satellite to ground image transmission bandwidth has been unable to meet the needs of the rapid growth of remote sensing image data.There are a large number of invalid data areas(cloud areas)in the satellite remote sensing images,and the cloud areas have less or even no usable information,which occupy a considerable part of the satellite-ground transmission bandwidth and data storage space in the process of image downlink,which will lead to a significant reduction of bandwidth utilization of effective data.Aiming at the above problems,this paper designs a set of remote sensing cloud images compression scheme,and implement it in FPGA for high-speed design,which can effectively improve the compression performance of remote sensing cloud images and effectively enhance the bandwidth utilization of satellite-ground transmission.The innovation points and main works of this paper are as follows.(1)Based on the analysis of gray histogram characteristics,texture characteristics and frequency characteristics of remote sensing cloud image,we proposed a cloud elimination scheme and a cloud quantization scheme based on local inner and outer edges.In the performance comparison tests with direct JPEG2000 compression,ADR+JPEG2000compression and LEC+JPEG2000 compression schemes,the PSNR values of the recovered image ground object areas of the cloud elimination scheme in this paper are improved compared with the other schemes,and the compression performance is effectively improved;the PSNR values of the recovered image ground object areas of the cloud quantization scheme in this paper are improved compared with the direct JPEG2000 compression scheme,and the information of the ground object areas covered by thin cloud area can be retained,and the overall compression performance is also improved.(2)In this paper,We completed the FPGA high-speed design and implementation of the overall scheme,the core processing is implemented by three-stage inter frame pipelining scheme and the overall speed can reach 100 MHz.In particular,the fast connected region marking algorithm in this paper only needs two traversal to complete the fast connected region marking.We designed and implemented the read-write control and arbitration module integrating the ping-pong operation and series-parallel conversion technology to control the DDR4,which can ensure the normal operation of the inter frame pipelining scheme,relieve the pressure of storage effectively and improve the system efficiency;We designed and implemented the morphological filtering expansion erosion algorithm by controling the Brams reasonably to realizes the horizontal and vertical organization of pipeline processing.We design a simple and feasible minimum weight finding scheme to construct Huffman tree,which can achieve fast and efficient Huffman coding.In order to verify the function and performance of the overall design,we designe different testbenches to simulate the function of FPGA design by modelsim.We carry out a large number of tests on the Alveo U250 board.At the same tine,we compare the results and speed between the hardware and the software.We come to the following conclusion: The resource usage of the FPGA high-speed design for this scheme is as follows: LUT 15936,LUTRAM1049,FF 13897,BRAM 136,URAM 0,BUFG 2,and the overall speed can reach 100 MHz.The average time consumption of each frame processed by software is 93.7 ms,and the average time consumption of each frame processed by hardware is 10.6 ms.The speed of hardware is 8 times higher than that of software.According to the experimental results and conclusions,the compression scheme of remote sensing cloud image based on FPGA has good performance and meets the real-time processing requirements on satellite,which has good application value.