This paper proposes a new background subtractionudmethod by a moving camera for the object detection.udKey points are firstly extracted and tracked. Fromudthe tracking results, spatial transformation relationships forudthe background scenes in consecutive frames are obtainedudwhile the current frame is warped to the previous imageudplane for the camera movement compensation. A codebookudbackground model is constructed and updated in an onlineudway by exploiting the full RGB color information, which isudused to distinguish the foreground/background regions. Bothudqualitative and quantitative experimental results show thatudthe proposed method outperforms its counterparts with audbetter performance.
展开▼
