Moving Target Detection In Complex Environments

Moving object detection is all important research field in computer vision. Related techniques have been broadly applied in video monitoring ,human-computer interaction , intelligent transportation system , military field and so on.The technology of motion detection belongs to the low but fundamental operation of computer vision. An effective algorithm plays a vital role to complete the follow-up work. Whereas, the dynamic background,the change of illumination,the moving shadow and the occlusion phenomenon may produce significant impact to the result of motion detection. How to get a high-quality moving object rapidly in complex environment has become the hot issues in the recent years.This thesis mainly focuses on the moving object detection methods in complex environment for further study. A certain amount of work has been done mainly in moving object detection and moving shadow elimination.Firstly,the classical algorithms of moving target detection are discussed and conducted the thorough research. Experimented on these means by real-time, spatial accuracy and temporal coherent evaluate objectively each algorithm and draw advantages , disadvantages and the suitable environment of each algorithm.According to the problems of classical moving target detection algorithms in complex scene, this paper proposes a motion detection algorithm based on a modified k-means clustering. Experiments prove this method not only can complete the task of moving object detection in complex environment, but also has good adaptability and robustness.Focusing on the disturbance of moving cast shadow, a semi-supervised learning method of multi-feature fusion based on support vector machine is proposed. Experiments show that this algorithm can eliminate the moving shadow effectively and improve the performance for the traditional methods.