Research On Object Detection And Tracking Under Complex Scene

Object detection and tracking is one of research hotspots in computer vision, and be widely used in intelligent video surveillance, human-computer interaction, and video conference. The main purpose of moving target detection and tracking is using computers to simulate the human visual perception functions, catch objects from image sequences, track the moving targets, identify and understand the behavior. This paper focuses on some key technologies in object detection and tracking. The research work may be summarized as follows:We present a new object detection algorithm based on background subtraction. First of all, this algorithm employee Mixture Gauss Model as the background model methods, and integrate improved inter frame difference to detect moving object. The problems of initializing background difficulty and updating background slowly are resolved in this method.Then we propose a new Mean Shift algorithm based on spatial edge orientation histograms, using space distribution and texture information as matching information. Meanwhile a Kalman Filter is used to predict the target's position. When the target is not occluded, we choose the predicted position of the Kalman Filter as the start point of the improved Mean Shift. Not only can it track the target accurately, but also, it can accelerate the process. When the target is occluded, we only use the Kalman Filter to predict the position of the target. And experimental results demonstrate that the proposed algorithm can deal with intricate conditions, such as significant clutter, partial occlusions, and it can track object efficiently and robustly.To overcome the shortcoming of single visual feature in complex scene, a tracking algorithm based on adaptive feature fusion mechanism for particle filter framework is proposed. The color feature and SIFT are utilized to represent the target and democratic integration is applied to fusion these two features. The result of experiment show that the algorithm adapts to the change of the object sharp variance, occlusion and clutter background.