Moving object Detection Based on Farneback Optical Flow

Moving object detection in an environment with changing illumination has been an interesting research topic recently. The optical flow method has been widely used for moving object detection as an effective method. However, the conventional optical flow method is constrained by illumination conditions. It cannot eliminate the negative impacts of illumination changes. In order to solve the negative impacts of illumination changes to the optical flow method, this paper proposes a moving object detection algorithm using the optical flow method combined with feature matching. The algorithm uses the farneback optical flow method to obtain the optical flow vector with noises. It uses the Canny operator to extract the moving object and the noisy contours from the optical flow. Then, feature extraction works to obtain multidimensional features. By weighted matching calculations of multidimensional features in the two optical flows, the extraction of the moving object from the noisy optical stream is completed. Experiment examples verify that the proposed method can detect moving objects accurately in the environment with changing light.