Skip to Main Content
This paper presents a new robust real-time system for multiple object tracking in traffic scenes. Different from existing object tracking systems this system utilizes an optical matrix range sensor which is mounted onboard on a mobile vehicle. The new generation of matrix range sensor acquires directly dense matrixes of range data from 3D environment at a frame rate up to 50 fps, and brings a new aspect of real-time computer vision processes. In this system the traditional region-growing method is applied to range images for the segmentation of objects, and a 2D Kalman filter model is designed to track objects on the ground plane. A new object association strategy is also proposed in the paper, which is capable to deal with object tracking robustly in case of merging and splitting. The presented tracking system is tested online under real traffic scenarios. Elaborate experiment results with real data are provided in the paper to evaluate the robustness and efficiency of the system.