A Novel Curb Detection Method for Advanced Driver Assistance Systems

摘要

Curb detection is an important research topic in environment perception, which is an essential part of Autonomous Land Vehicles (ALV) and Unmanned Ground Vehicle (UGV) operations. Also, curb detection is specifically a component in Advanced Driver Assistance Systems (ADAS) that is assisting drivers in parking their cars without hitting the concrete curbs/parking blocks. In this paper a novel curb detection algorithm is developed by using Canny edge detection along with Hough transform. A CMOS camera fixed at the bottom part of the mesh grill, which is at the center of the front bumper, is used to acquire images. The camera is interfaced with a computer running an algorithm based on the Canny edge detector and Hough Transform to detect the edges of the curb. The developed program is also capable of measuring the distance from the car front end to the curb. Distance calculations are established based on a calibration technique, where rows within the camera image frame are mapped to corresponding distances in front of the vehicle, assuming a flat terrain. The method is able to detect curbs robustly in typical static and dynamic environments in real vehicle experiments.