Research On The Segmentation Of Moving Object In Video Sequence Based On Embedded System And Its Implementation

Video segmentation is a key technology in a new generation video coding, video browsing, internet multimedia interacting etc. It has been applied to the domain of robot vision, inspecting system, video access, product detecting, videophone and video conference abroad. However, the application is mainly aimed to PC whose computing capability is relatively strong and memory space is relatively loud, but it needs embedded system in many cases. So, the research on the video segmentation based on embedded system has become very significantly.Firstly, the conception of embedded system is introduced and the capability of Windows CE is analyzed. The harvest and the problem of the video segmentation at present are discussed. Several algorithms that based on background model are analyzed. And the reason that those algorithms cannot be applied to embedded system is introduced.Secondly, an algorithm based on embedded system is presented. The background model is established by the strategy of average value, and updated using a dynamic matrix, the moving objects can be segmented by difference image method, the pixels of global lighting change and moving background can be updated to background adaptive. The algorithms whose time complexity is 1. 675×H×V×T and space complexity is 6×H×V can fully meet the requirement for the embedded system.Finally, a frame of application system is introduced in this paper, and the algorithm is realized in the Windows CE. The experimental results demonstrate that the moving objects in the case of single object, multi-objects and moving background can be segmented well and truly, and it is effective to track video moving objects in embedded system.