Journal of Engineering Research

Aiming at the problem of roadside extraction in passable area, a method based on artificial feature extraction is proposed by using only three-dimensional LIDAR point cloud. The method obtains the accessible area of the road by detecting the roadside points. For roadside detection, a moving base-point and base-axis method is proposed, which selects a new base-point and base-axis at all times, so that the base-axis and base-points can flexibly adjust the direction and position to improve the accuracy of the classification and extraction of non-straight roadside points. And the least squares method is used to the roadside point collection curve fitting to obtain the smooth road extension trend, which realizes the stable output of the plane roadside curve. In addition, in order to solve the fluctuation problem between frames, Kalman filtering algorithm is used for tracking and identification of road edges, and by constructing the prediction and observation model, the stable output of roadside points is realized, which ensures the reliability of the system under different vehicle speeds and road conditions.

Roadside extraction; LIDAR; Artificial feature extraction; Roadside detection; Curve fitting; Extended trend