Research And Implementation Of Moving Target Tracking Algorithm Based On Multi-core DSP

With the development of the society,video surveillance system has gone into the tens of thousands of households.Intelligence is the main development trend of the surveillance system.And achieving the tracking of the moving target in the surveillance system is an important manifestation of intelligence.Taking intelligent video surveillance as a research focus,this thesis firstly designs the implementation platform of target tracking algorithm-the specific video surveillance system based on multi-core DSP.Then it studies the movement target tracking algorithm.Finally,the algorithm is realized on multi-core DSP monitoring platform.The focus of this paper is as follows:(1)Taking TMS320DM8168(referred to as DM8168)multi-core processor as the core,the specific video surveillance system is designed to achieve the video capture,processing,display and network output and other functions.The characteristic of the system is that the video streaming framework is designed according to the actual needs,and the video is made TILER transform,OSD logo added,mosaic spliced and other processing.Besides,the streaming media server program is designed to achieve the function of the video network output.In this design,the stability of the system is improved by adding AVS function,and the scalability of the system is enhanced by designing the PCIe driver.(2)In this thesis,the feature extraction algorithm is an important part of the target tracking algorithm.By comparing the various feature extraction algorithms,this thesis chooses the SURF algorithm with performance and efficiency to extract the characteristics of the target.And the SURF features of the target are matched by a combination of multiple matching algorithms.A reasonable simulation flow is designed in the overall testing of the target tracking algorithm.A solution to the common problems such as target morphological change,target occlusion and target shadow influence in target tracking is presented,and the simulation results are analyzed.(3)According to the characteristics of DM8168 multi-core surveillance platform,the implementation process of target tracking algorithm with multiple cores involved is designed,and the algorithm is implemented according to the process.As real-time requirements are very high after the implementation of the algorithm,the algorithm is deeply optimized according to the characteristics of DSP,so that the speed has been improved.The implementation effect is validated by testing the target tracking algorithm.(4)In order to better guarantee the performance and speed of the algorithm,the SURF algorithm which takes the most time in the target tracking algorithm is implemented on TMS320C6678(referred to as C6678)platform.The entire task in implementation is divided by the way of dividing the picture so that each core deals with a sub-task.The correctness of this division method is proved by the analysis of the processing results.And the speed of the algorithm has been greatly improved.The exploration to the way of the realization of SURF algorithm on C6678 platform has laid a foundation for the realization of the follow-up target tracking algorithm on the whole platform of DM8168 & C6678.