The Research On Moving Objects Detection For Complex Background

Moving object detection is one of the important research directions in computer vision, and also is the core of intelligent video surveillance systems. However, moving object detection algorithm is easily disturbed by external factors in dynamic environments, which makes it difficult to meet the needs of high-level visual processing systems. The research of moving object detection in complex environments is carried out in this dissertation. And we also summarize and analyze the definitions of moving objects, the existing moving object detection algorithm and the background disturbance factors. We propose the EVibe (Enhanced Visual Background Extractor) moving object detection algorithm that can avoid the disturbances of the dynamic background. To refrain from the influences of shadow, we propose a shadow detection algorithm based on a combination of chromaticity and gradient and add it into the EVibe algorithm as a shadow removal module. The results show that our algorithm improves both the detection rate and the recognition rate. Our major contributions are:(1) A systematic summary and analysis of the moving object detection problems under the complex background. Analyzed the definitions of moving object, rigid moving object and non-rigid object. Proposed the essential attributes of each kind of moving object. Summarized and analyzed the disturbance factors that disturb the moving object detection and the classical moving object detection algorithms. So that we can choose suitable detection algorithm in practice.(2) A shadow detection algorithm is proposed based on a combination of chromaticity and gradient. Analyzed the principle of shadow formation in order to avoid the shadows’ effect on the moving target detection algorithm. The shadow detection algorithm was proposed on the basis of a combination of chromaticity and gradient, which has been used in the object detection systems and an improved result has been achieved.(3) The EVibe detection algorithm is proposed. To reduce the affection of the disturbance and confounding factors under the complex background in the detection algorithm, we improved the Vibe(Visual Background Extractor) algorithm and propose the EVibe algorithm. EVibe algorithm provided a better solution to decrease the affection of disturbance factors such as light, shaking trees, shelter and camera shake.