Design Of Video Surveillance And Pedestrian Detection Systems

With the development of China's economy and the emphasis on the construction of safe cities and intelligent transportation systems,the demand for video surveillance systems has steadily increased,and the requirements for surveillance systems are high-definition and intelligent.At present,the market is still dominated by ordinary monitoring systems,with low compression ratio,high cost,and only ordinary video recording functions,so the intelligent monitoring system has become a hot research topic.Thesis designs an intelligent monitoring system with pedestrian discrimination and high compression.The background and meaning of the video surveillance system are understood,and the video compression coding technology and pedestrian detection algorithm are studied.The system consists of two parts: the video client and the server.The monitoring server uses Hi3516 A as the control chip,and the client is implemented on the PC side,which realizes functions of video data acquisition,compression coding,video transmission and real-time pedestrian detection.The design of video surveillance and pedestrian detection system has been completed.Thesis has completed the following research work:(1)According to the development trend of video surveillance system and the requirements of the project,the overall design scheme of the system is proposed,and the related hardware and technology are understood.(2)The pedestrian detection algorithm is studied,and a fast pedestrian detection method based on hybrid Gaussian background model and HOG-LBP feature is built.The moving target is extracted by background modeling,and the edge detection is performed by Sobel algorithm.Finally,HOG is used.LBP feature and SVM classifier detect image,this method effectively improves the detection speed,can reach 20 frames per second,the accuracy rate is above 94%,and can quickly and accurately detect pedestrians.(3)Designing the overall scheme of video surveillance,and doing a detailed study on the technical difficulties,including the new generation of H.265 video compression coding technology,streaming media transmission protocol and video display technology,Linux system environment construction including cross-compilation environment,Opencv Transplantation,etc.(4)Research on the media processing software of Hi3516 A,and develop the acquisition,processing and compression coding and transmission of video images on MPP.In this system,the monitoring client is developed and applied on the PC.Ffmpeg is used to decode the H.265 encoded video,and the video image is displayed through DirectDraw,which realizes video decoding,video display and video playback control.(5)Completing the construction and testing performance of video surveillance and pedestrian detection system,the results show that the H.265 compression efficiency of this system is increased by about 50% compared with H.264,and the video delay is between 300~350 ms.Pedestrian detection algorithm With an accuracy of more than 90%,it has good real-time performance,stability and reliability,and meets the requirements of system design.