A Robust Curb Detection Algorithm Based on Lidar Sensor
Yongjian Zhu,Shixuan Tian,Xiubo Ye,Benyuan Wang,Yi Huang,Kun Tang,Frank Jiang
DOI: https://doi.org/10.2139/ssrn.4292738
2022-01-01
SSRN Electronic Journal
Abstract:Curb is a key component of the structured road, so curb detection is an important task in environmental perception. To correctly extract curb edges of the road around the self-driving vehicle in multiple traffic scenarios, this paper proposes a new road edge detection method based on 3D LIDAR sensor. We design a new road-edge point extraction algorithm combining multiple spatial features with loose thresholds.In which each scan line of the LIDAR is processed separately, and the curb candidate points are extracted while being able to discriminate the left and right edges. Then a curb candidate point filter is proposed to eliminate false detection points by combining density filtering and RANSAC filtering methods. In result, the comprehensive experimental evaluation shows that our algorithm can effectively identify the curbs of both straight and curved roads in the presence of obstructions such as vehicles and pedestrians, and the average Precision, Recall, and F1 Score are above 80%, which verifies the accuracy and robustness of our method.