| Video surveillance system is a crucial issue in the field of computer vision,pattern recognition and artificial intelligence, and it has broad application foreground for safety monitoring, intelligent transportation and military navigation. Moving object detection technique is an important component of video surveillance system, its detection results directly affect latter target location, identification and tracking, as well as comprehension and description of movement behavior. Large numbers of researchers devoted themselves in the area and have already achieved many progresses. The dissertation studied moving object detection algorithm of video surveillance system based on the current conclusion.This paper first introduces basic knowledge of digital image processing, and according to the actual condition, contrast verification several normal image processing methods, such as noise filtering, color space selection and transformation, image edge detection, and mathematical morphological processing, which provide a foundation for algorithm research of moving object detection and shadow suppression.In moving object detection aspects, aiming at background subtraction, analyses and studies on usual background modeling method, and a kernel density estimate background modeling based on key frame sampling scheme is proposed. Taking advantage of average background of the interval images sequence for preprocessing of samples, and according to pixel similarity theory, choosing those samples that have critical background information for building background model. Using threshold segmentation to transform grey level image into two black-white value image, then images including moving information are filtered by morphological filtering to eliminate background noise, the object contour is abstracted. Meanwhile an updating mechanism combining timing updating of whole samples with real-time selectivity updating is also applied for auto-updating background model. The algorithm has the advantage of fast speed and high precision, and it can effectively deal with the problems of the building and update of background, background disturbance, illumination changes etc.In moving shadow suppression aspects, the accuracy and integrity of objects detection depend on moving shadow suppression algorithms to a great extent. Most of the current approaches proposed are mainly based on shadow optical properties. These algorithms are simple and easy to be realized, but they focus mainly on a certain aspects of shadow properties, without generality. In this paper, a shadow detection algorithm combining geometry model with shadow properties based on 8 direction Sobel operators and gauss kernel density clustering is proposed by analyzing the current algorithms of shadow detection. Firstly, using 8 direction Sobel operators to extract shadow image edge after shadow existence judgement, and obtain sketchy shadow regions by morphology processing, then a gauss kernel density clustering algorithm for luminance gain image are employed to detect the shadow based on the characteristics that luminance gain has a very small torque fluctuation between shadow and background region. Thus moving objects and shadow segmented completely.In practical application aspects, a moving vehicle detection system is designed based on IST. System designer idea and equipment composition are given, and a modular design is presented according to actual condition, including working principle of module and system design flow.In this thesis, according to the deep research of detection problem on moving object, my ways is put forward to solve them and test them through experiments. The result from the experiments show that the proposed moving objects detection algorithm can fast detect the moving region completely, and it is robust against noises and disturbance of the background. Shadow detection algorithm based on 8 direction Sobel operators and gauss kernel density clustering can effectively suppress shadow of moving objects. |
PDF Full Text Request