free_throw

題目:在壅塞環境下的多人追蹤

Multiple Human Tracking in Cluttered Scene

Multiple human tracking is an important research topic in computer vision. When an occlusion occurs, the problem becomes more challenging. In this thesis, we propose a tracking system to track each human object and solve the occlusion problem in a cluttered scene. We use color and motion, the two significant features, to distinguish and track different human objects. We classify each pixel to which human object it belongs by its relative position about an object and its color. Combing pixel classification of each frame and optical flow, we can track each object independently. When an occlusion happens, we analyze the color distribution of the occlusion group to differentiate each object in an occlusion. By calculate distances between each objects, we can determine whether an object separate from the occlusion or not. For an object leave an occlusion, we can track it correctly afterward.