Detection And Tracking Of Moving Object In Vision

This thesis discusses visual tracking problem. It involves in the following several topics, moving objects detection, tracking the objects and camera calibration. Visual tracking is one of the most important branches in machine vision, and is the basis of some advanced machine vision. It has great applications such as video surveillance, virtual reality, moving object capture, intelligence transportation and military guidance.On the research of moving objects detection, primary algorithms of segmentation are introduced, and then an improved algorithm based on adjacent frame difference is proposed. Statistic characteristic of the background image is investigated. An adaptive-refreshed background model is proposed. Noise problem and shadow problem of detecting objects are investigated. Experiments are given to show the validity of algorithm.On the research of the tracking objects, different tracking models and current research aspects are discussed. An adaptive kalman filter-tracking model is proposed to track the objects maneuverable motion. A matching algorithm based on color histogram is used in the objects matching under Bhattacharyya space.On the research of the camera calibration, classic calibration algorithms and self-calibration algorithms are introduced. By means of a self-calibration, mapping from object of 2D to 3D is realized. It is the base of further topic of object capture.