A New Robust Multiple Pedestrians Detection and Tracking System

In Phase I, we have developed new pedestrian detection and tracking techniques based on stereo images. Our system comprises a number of key modules including (1) disparity layered thresholding method that can be used to extract the potential pedestrian(s) layer by layer; (2) pedestrian detectiona and discrimination using 3D convex feature that can be used to detect the pedestrian in each disparity layer; (3) continuous pedestrian association that correlates the same pedestrian in subsequent image frams; and (4) pedestrian motion estimation. We have tested our Phase I system using stereo images of pedestrian walking cross street intersections under various weather conditions. The test results have shown that our system can effectively detect and track multiple pedestrians. Our primary goal in Phase II is to develop a commercial product of pedestrian early warning system with both software and hardware. There are three hardware components in our Phase II product: (1) IR LED stereo camera, (2) eletronic processing unit, and (3) traffic controller. Our disparity estimation algorithms and pedestrians detection and discrimination algorithms will be deployed in the electronic processing unit. We have teamed with two companies for the product development in Phae II, sales and marketing in Phase III.