Research On Image Data Compression Based On Embedded Platform

Multi-node sensor fusion is an important trend in the development of industrial Internet,but the problem that follows is the contradiction between the expansion of data volume and the limitation of communication bandwidth,and the real-time requirements of industrial detection are also difficult to fulfill.Compressing the data at the upstream port where data are generated to reduce the pressure on the communication system is an effective way to solve this problem.Based on this problem in the actual engineering project,our team in this thesis has carried out related research,built a compression system based on the embedded platform and achieved the expected design goal.The specific work is as follows:(1)After analyzing the two main types of classic lossless compression algorithms,LZO algorithm suitable for this system was selected,Applied the open source library mini LZO,and experimental tests were conducted to verify its feasibility and reliability.Comprehensive compression ratio,time-consuming and other indicators were determined.Applicability in this system.(2)A compression system and a simulation experiment environment based on an embedded platform were built.Performance tests on them were conducted.The LZO algorithm was transplanted to STM32 to build a compression system based on the embedded platform.At the same time,the experimental environment is simulated by the host computer software written based on Labview.Finally,the performance of the built system is analyzed and verified.(3)An optimization method is proposed based on the spatial redundancy of image data,and two solutions of lossless compression and lossy compression were provided,which could be selected for different precision demand scenarios.Finally,the proposed method was tested and verified.The built system can compress an image file to about 49% of its original size in 19.3ms.The optimized algorithm compresses the image to 38.16% in the lossless state,and can compress the image to 19.5% with an error of 0.29% in the lossy state,which can achieve the expected performance index.