Robust Curb Detection from Multi-frame 2D Laser Range Finder Data

摘要

Curb detection is an important research topic in the unmanned ground robot.In this paper,a novel curb detection method using a 2D laser range finder in an urban environment was presented.In the proposed method,firstly,2D sequential laser range finder data were aligned using a Dead-Reckoning(DR) and GPS localization method based on extended Kalman filter.Secondly,curb points were extracted from multi-frame laser data based on spatial-temporal correlation analysis.Finally,the straight and curved curbs were detected by coordinate axis conversion least square linear fitting method and piecewise weighted least square fitting model respectively.The proposed method can detect both raised and sunken curbs robustly in campus environment.The proposed method has been verified in a real robot platform.