Pedestrian and Bicycle Detection and Tracking in Range Images

This paper presents a real-time algorithm for detecting and tracking bicyclists or pedestrians using a laser device. By processing the sequence of the range images, the algorithm outputs trajectory and speed of each object during the period when he is in the detection region. The whole algorithm consists of two parts, which are the object detection and the object tracking. In the former, the multi-level thresholding method is combined with the Iterative Selforganizing Data Analysis Techniques Algorithm (ISODATA) to implement object segmentation. In the latter, Kalman Filter is applied to recognize and track moving objects. Experimental results demonstrated this algorithm is effective in object recognition and tracking, as well as robust in the applications.