Research On Moving Object Detection Algorithm In Intelligent Video Surveillance System

With the rapid development of the discipline of computer vision, pattern recognition, computer technology, communication technology, the intelligent video surveillance system has become an important means of security defense and has great application value.And moving object detection, refers to automatically remove the background from the video and extract the moving objects that people interested in. Moving Object Detection is the foundation and core technology of the intelligent video surveillance system.However, because there is usually illumination change, the interference of background object movement such as shaking leaves and shadow in the scene, so making it difficult to establish a real-time and accurate moving object detection method.Firstly,research moving object detection algorithm in camera static condition.Includes the following aspects:research the existing motion object detection methods’theory and the advantages and disadvantages, Including frame difference、time-mean、time-median、 single gaussian model、gaussian mixture model.Then, propose an improved method based on Single gaussian mode approach and an improved method based on Gaussian mixture model approach,including improved Gaussian mixture model initialization phase and the update mechanism, After light mutation re-initialization of the model or combined with the gaussian mixture model and frame difference method to improve the adaptability of the light mutation, adaptive change the number of each pixel’s gaussian distribution and the use of block-based moving object detection method to improve the speed of the algorithm.Than, analysis the impact of the shadow to the moving object detection algorithm as well as the characteristics of the shadow, the common shadow removal methods, research and realize the real-time shadows removal method based on HSV color space.Finally,Realize a system to filter background images in the video sequence.