Research And Implementation Of Embedded System Image Foreground Detection Technology

Foreground object detection is usually the basis of the intelligent videosurveillance system behavior analysis and understanding, the foreground detectionresult is good or not directly affect the reliability, stability and robustness of the entirevideo system. With the development of embedded software and hardware technologyand the computer vision technology in recent years, based on the embedded intelligentvideo surveillance systems are increasingly favored by the market for its high real-timeperformance and stability, the research and development of a highly intelligent videosurveillance system, and has a practical application of the system to achieve the producthas become nowadays one of the most popular topics. This dissertation has somein-depth research on foreground detection technology and specific detection algorithmis ported to the embedded platform to achieve real-time detection of the foregroundobject.First, the selected research background, research significance of the issues，involved in the intelligent of the monitoring field, foreground detection technologydevelopment status and development background of embedded DaVinci technology aredescribed in this dissertation. The typical foreground detection algorithm and a varietyof methods to build a background model are studied detailedly. The basic ideas andprinciples of these algorithms are described, and through the comparison of theexperimental results, the advantages and disadvantages of these algorithms are analysedand discussed, as well as the application of occasions, according to the transplantalgorithm selection.Then, the algorithm ported target platform, Netra TMS320DM8168, mainlyincluding the DM8168structure of hardware and software development frameworkMcFW, are described in the dissertation. The hardware structure is introduced, includingthe ARM, C674x DSP and video processing HDVPSS subsystem features. Softwaredevelopment framework McFW is an unique software architecture of DM8168, theinternal structure link interfaces as well as the link creation order rules are introduced.The building of DSP embedded software development environment is introduced, mainly including the installation and configuration of DM8168SDK, Network FileSystem NFS and tftp server.Finally, according to the characteristics of DM8168software architecture, theoverall monitoring system is divided to modules, the algorithm of three differentialmethod for foreground detection is ported to DM8168, mainly including the algorithmimplementation in the DSP end as well as the call in the application side, and builds theexperimental platform based on the DM8168development board, the portable algorithmvalidation, experimental results show that the detection algorithm is good, and toachieve a real-time video data collection, testing and processing requirements.