Moving Object Detection from Video with Optical Flow Computation

作者：Zhan Wei^*; Yang Junkai

来源：Information-An International Interdisciplinary Journal, 2012, 15(10): 4157-4164.

• Moving object detection
• Optical flow
• Image pyramids

Moving object detection is an important branch in research on computer vision. Traditional method of motion detection is image difference though sequential two frames or background which result is apt to be affected by brightness changing of image, so in this paper, a kind of moving object detection algorithm based on optical flow is purposed: first and foremost, using image pyramid optical flow estimation algorithm, calculate optical flow image of two consecutive frames and figure out threshold value of current optical flow image; then via threshold value get in first step, a binaryzation image of current frame came into being; third, using closed operation which is a kind of morphological transformation on binaryzation image, appearance image of moving object is obtained. Take moving vehicle video files including moving vehicles for experiment, using above algorithm, moving cars are separated efficiently and accurately from background.

• 出版日期2012-10
• 单位北京大学; 中国地质大学（北京）