A Roadside Monocular Vision Method for Target Ranging and Speed Estimation Using 3D Object Detection

With the development of intelligent transportation systems, roadside video perception is expanding from traditional detection and tracking to higher-level measurement outputs, such as inter-target distance and target speed. These measurements are important for traffic flow estimation, accident warning, and timely responsibility assessment after incidents. Monocular cameras offer low cost, easy deployment, and wide coverage, but recovering reliable 3D scale from 2D pixels remains difficult; therefore, achieving stable monocular ranging and speed estimation under practical deployment constraints is still challenging. In this paper, we propose a roadside monocular vision–based method for target ranging and speed measurement. By combining a 3D object detection approach with coordinate transformation, the method outputs real-world target coordinates and speeds in the scene. We conducted several experiments, and the results show that our method can obtain target spatial coordinates and speeds in surveillance settings with reliable performance.