Moving Target Detection Technology Research And Implementation Under The MPEG-4 Compression Domain

MPEG-4 video is widely used in security monitoring and intelligent traffic areas. This thesis studies moving target detection under the MPEG-4 compression domain, and its main point has three parts.1. MPEG-4 codec method research.Research for different video object coding method. motion compensation coding based onTwo values shape coding and arithmetic coding. Discrete cosine transform, quantification, DC&AC forecast and variable length coding constitution the texture coding. And the main principle of motion coding method. Analysis of MPEG-4 video decoding process2. Research on moving object detection.Research the DCT coefficient method and motion vector method, Combined with the advantages and disadvantages of both,put forward the improved moving object detection algorithm, based on the combination of the motion vectors and DCT coefficients. The algorithm uses motion vector and texture informations in the compressed stream, through iterative filtering, four neighborhood growth, preliminary determine the moving target. Correction the moving target edge use the consistency of spatial feature vector. The characteristic of this algorithm is deal with to make full use of the information contained in the compressed stream and combination the main reason for static background noise. Compared with only using motion vectors and DCT coefficients detection, the performance of the algorithm has been greatly improved and deal the real-time processing of video moving objects in the stream.3. Moving target detection system design and implementation under the MPEG-4.With the idea of software engineering design and implementation of moving target detection and analysis system.The hardware system is made of video camera, HaiKang hard disk recorder,and computer. Developed the moving target detection software in Microsoft’ Visual Studio 2010 development environment on the basis of HAIKANG SDK, OPENCV and FFMPEG. The software includes configuration, image acquisition and motion target detection and analysis of three parts.. After testing, the software realized the function and has a good stability.