Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research on Moving Object Detection and Tracking in Complex Condition

Author LiuRui
Tutor SunHan
School Nanjing University of Aeronautics and Astronautics
Course Applied Computer Technology
Keywords Moving target detection and tracking Multi - feature fusion modeling Particle filter Line scan DSP transplant Optimization
CLC TP391.41
Type Master's thesis
Year 2009
Downloads 152
Quotes 1
Download Dissertation

Moving target detection and tracking is a key topic in the computer vision -disciplinary research . Has a wide range of applications such as video surveillance, security suspects under strict surveillance , weaponry . Through the joint efforts of the researchers at home and abroad over the years , moving target detection and tracking technology has been developed by leaps and bounds , and made ??a lot of ground - breaking research . But in actual complex environment of the scene , the illumination changes , object occlusion , the presence of shadow interference , deformation effects , presented new challenges to the technology . To solve these problems well , and to achieve the requirements of practical applications , the need to design real-time and robustness of the new algorithm both . This paper in-depth study of certain core algorithm , and made ??a new attempt on this basis , this paper is as follows : depth study of the classical methods of detection and tracking of moving targets at home and abroad , and experiments to understand the advantages and disadvantages of each and suitable application conditions. On this basis , the proposed combined modeling based on multi-feature target Mean-Shift and Particle Filter tracking algorithm . The goal modeling combines general target characteristics as color , edge , texture , and also draw on the concept of structural similarity (SSIM) image quality evaluation in the field of research , the structural information extracted target characteristics as the fourth fusants . Improved multi- feature fusion method , the robustness of the target model . Final tracking of moving targets in MSPF (Mean-Shift Particle Filter) improved tracking framework . The algorithm to track accurately verified by experiments . Proposed a target extraction algorithm based on line scan , the algorithm for progressive scan image , extracted directly target chain code tracking , do not need the full map . The algorithm combined in the development of embedded DSP DMA double buffering technology , can greatly reduce the overhead of algorithm time . The multi - feature fusion the MSPF target tracking algorithm transplant to the the ADSP BF561 development platform , a number of issues to solve transplantation in image format conversion . Combined target extraction algorithm based on line scan , optimize the realization of the target tracking algorithm in the platform . Transplant based on the C language and assembly-level code optimization , to further enhance the efficiency of algorithm execution , and ultimately to meet the requirements of real-time target tracking applications .

Related Dissertations
More Dissertations