H.264 Encoding Algorithm Improvement And Embedded Monitoring Application Research

Video surveillance is an important part of traffic management system.The continuous working mode generates a huge amount of data,which causes great pressure to data transmission and storage.Massive amount of raw data,limited bandwidth and hardware storage space together make it extremely difficult to transfer,process and store monitoring data in real time.Therefore,this thesis aims to study the video surveillance system,and improve of coding efficiency of H.264 algorithm,so as to adapt the application of real-time road monitoring on low-cost embedded system platform.The main work of the thesis includes.Research on the framework of H.264 coding algorithm and all parts of the technology are carefully carried out,on this basis,the inter-frame prediction motion estimation algorithm and the intra-frame prediction algorithm are improved.In the inter-frame predictive motion estimation algorithm,the asymmetric cross-shaped search template,5×5 square search template,multi-level hexagonal search template,hexagonal search template and search window in the UMHexagon S algorithm are improved for the direction of road traffic monitoring applications,so that they are more in line with the monitoring road vehicle motion characteristics.Experiments under the JM model show that the inter-frame prediction motion estimation time and coding time are significantly reduced while maintaining the signal-to-noise ratio and coding bit rate basically unchanged.For the intra-frame prediction algorithm,the Intra16×16 and Intra4×4 patterns of the 16×16 macroblock are firstly predicted based on the improved threshold determination method to narrow the search range.Then,the predicted patterns are further narrowed down based on the texture direction for the predicted patterns.Experiments under the JM model show that the coding time decreases significantly while the signal-to-noise ratio and coding bit rate remaining basically unchanged.Superposition experiment of the two improvements under the JM model are carried out,and the test results show that the encoding time of the video decreases by about 24.48%,while the signal-to-noise ratio improves by about 0.07%and the encoding bit rate improves by about 1.43%.An embedded real-time video surveillance system is built and implemented,in which the Cortex-A53 microprocessor is selected as the core of the system acquisition terminal.The system software is built,and the improved H.264 algorithm is embedded into the video surveillance system.Linux is used as the operating system of the hardware terminal,and Nginx is selected as the video server to realize the functions of video acquisition,encoding,saving and transmission.The display terminal uses a PC to receive the video encoding information from the capture terminal,and uses Qt and FFmpeg decoding technology to build a program to decapsulate,decode,play and store the received information stream.The built video surveillance system is tested,which can meet the expected requirements and the application requirements of thre traffic video monitoring.