Moving Shadow Detection Suppression Algorithm And Implementation Of Intelligent Video Surveillance System

Along with the rapid development of the network technology, video processing technology and the upgrading of software and hardware in embedded device, the embedded video surveillance systems are very popular in the world. After the analog, digital, mixed phase, with the security industry, the video surveillance system has also developed inevitably toward the network, the digitizationand the intellectualized direction.As an advanced video application ,intelligent video surveillance system can monitor accidents according to the customer's requirements, and apply the quickest and best way to alarm and provide useful information in order to more effectively assist staff to deal with crisis. However, due to the complexity of the physical environment as well as limitations of the system's function and performance, many video surveillance systems are neither time nor accurate in various degrees and failed to reduce the phenomenon of false positives and omissions, so these problems cause the entire system's safety and practicality degrade, meanwhile the moving shadow has become very difficult to deal with in intelligent video surveillance.The moving shadow detection is the key technologies in the intelligent surveillance algorithm, it's because of the same behavior feature of the moving shadow and moving objects, simultaneously the shadow may also cause to connect some different objects to one, causing errors in target detection and tracing. However, under the highly request of real-time ability and hardware quality in an embedded intelligent surveillance system, an appropriate shadow detection and suppression algorithm (SDSA) has not yet been proposed to solve the problem.This paper main research SDSA and it's realization in the intelligent surveillance system. First, the development of the shadow detection and suppression algorithm at home and abroad were discussed and analyzed. After that, we introduced the preprocessing algorithms of SDSA, the paper introduced several kinds of commonly process in SDSA based on forming shadow and illumination's property, then pointed the advantages and disadvantages by comparing each algorithm. According to the characteristics of an intelligent surveillance system, this article improved the HSV-SDSA in the YUV space, and optimized the code of this algorithm. Finally, measured the performance of this algorithm by some AVI videos, and proved the validity and real-time ability of this improved SDSA in target detection and feature extraction through the experimental results. The improved real-time SDSA proposed by this article had laid the good foundation for the following algorithms of target tracking, target recognition and target analysis.