Automated driving is the development trend of vehicle driving, and laser LiDAR is the high-precision positioning sensor that automated driving relies on most. Laser positioning mainly depends on the relative orientation from the detected environment features to the self-vehicle. However, when there are multiple detected features that carry different orientation information, all the data are redundant. In order to improve the vehicle’s positioning accuracy, two laser point evaluation models were put forward. First, based on the plane area analysis method, a spot error model was proposed, and the error distribution with the incidence angle and scanning distance was obtained. Second, the laser point’s position approximately obeys the Gaussian distribution, the laser’s stereo error ellipsoid model was established, and the laser point’s probability volume was calculated through the probability integration. Third, the point cloud features were extracted by a proposed roughness method, the point accuracy was evaluated by the proposed spot and stereo error models, and the points’ relative weights were recalculated in the laser signal’s positioning process. Finally, in order to verify the laser point’s evaluation positioning method, the simulation and experimental verifications were conducted. The results show that the evaluation method based on the error ellipse models can improve the laser positioning accuracy effectively.

No abstract