A Retrograde Vehicle Detection Method Based on the Optical Flow Field
Youting Zhao, Xiying Li, Zhenbo Wang · 2009
Vehicles generally run at high speeds on the highway, so the presence of retrograde vehicles on the road greatly increases the probability of traffic accidents. In view of this situation, we introduce a new macro-method to detect retrograde vehicles based on optical flow. Our approach mainly uses pyramids with the Lucas-Kanade motion estimation method, which greatly reduces computational complexity. Since the flow of traffic is regular and physically measurable, its optical flow field is also orderly and regular, allowing the optical flow vector to accurately reflect the velocity of traffic flow. Experiments show that retrograde vehicles on the highway can be well detected by finding abnormal changes in the optical flow field.