Segmentation Study Of The Extraction Algorithm, Based On The Goal Of Intelligent Video Surveillance System

Intelligent Video Surveillance System is a new intelligent digital video processing system. It uses normal analog camera to collect real-time field scenario (eg. communities, traffic situations, people flow in public places, product lines in industries, etc), which is transmitted to embedded high speed DSP real-time image processing board, where target (person, object or car) is extracted from background and compared with prescribed rules in order to judge whether the target has violated the rules. Once violation event happens, alarm information is generated automatically, and automatic tracking is initialized in multiple-camera based networks. This system has the function of intelligent real-time active surveillance, which can avoid off-line manual processing afterwards. DSP makes use of computer video surveillance to automatically analyze image sequences from video camera, and then gives judgement on the behaviors of target so that target localization, recognition and tracking in dynamic environment can be realized. Therefore, this system can not only achieve daily management but also provide real-time abnormal situation processing, reacting and real-time active surveillance.Based on the existing intelligent video surveillance system, this paper, firstly, improves the hardware design of intelligent video surveillance system, including the achievement of embedded high-speed DSP real time image processing board based on TMS320DM642. Secondly, we design video image storage and playback module software based on JPEG codec algorithm, which is achieved in the DSP-side and PC-side software. This design enriches the intelligent monitoring system and expands the market scope of application of the system. Thirdly, this paper investigates the image segmentation algorithm, especially the algorithm on image segmentation and moving object segmentation and extraction, which focuses on the object segmentation and extraction algorithm based on non-boundary dynamic contour and the two-step segmentation algorithm in video monitoring system. The simulation results of these algorithms and also the improved solutions are given. Finally, it is introduced that the optimization methods of DSP system, mainly the optimization methods of DSP systems and image processing algorithms in the process of its achievement in the DSP system; and then the results of optimization experiments are given.This paper involves four parts: the hardware design and improvement of the intelligent video surveillance system; software design and modules development; algorithm research and optimization of DSP system, where the system design and algorithm research on object segmentation and extraction are the focus. With the establishment of image and background, algorithm research has paid more attention to the contour segmentation and extraction of multi-objects and a variety of irregular outline of the targets.