Ranging accuracy difference correction among channels of a vehicle-borne road detection LiDAR
Tao Yang,Wei Yan,Jiancheng Lai,Yan Zhao,Zhixiang Wu,Yunjing Ji,Chunyong Wang,Zhenhua Li
DOI: https://doi.org/10.1016/j.optlastec.2023.110477
IF: 4.939
2024-02-01
Optics & Laser Technology
Abstract:Ranging accuracy is an important index to evaluate the performance of a LiDAR, which determines the LiDAR's resolving ability. In order to cover a wider detection range, there are directional differences between the channels of a vehicle-borne LiDAR usually, which will cause the lasers in different channels to act on the target at different incident angles, which will have impact on the ranging accuracy in each channel. Based on the working principle of a LiDAR, we establish the echo signal model of each channel, and deduce the detection Probability Density Function (PDF) of each channel. After that, we use Monte Carlo method to simulate and analyze the influence of the channel difference on the Systematic Error (SE) and Root Mean Square Error (RMSE) of ranging results, and correct the data by analyzing the characteristics of the statistical distribution in each channel. Then we design a corresponding experiment to verify the simulation results. The experimental results show that when the channel's pitch angle increases, the statistical distribution of ranging results deviates from the normal distribution gradually. The range of statistical distribution becomes wider, and presents an asymmetric distribution trend, and the lower the Signal-to-Noise Ratio (SNR), the more obvious this trend is. We find that the distributions of the ranging results' SE and RMSE both fit the exponential function well. Based on this, we propose a data correction method to correct the variation of ranging accuracy caused by channel differences. The LiDAR is placed at a height of 2 m and the vertical field of view (FOV) is −25°∼ +15°, and we set up three obstacles (0.5 × 0.5 × 1.8 m 3 ; 1 × 1 × 1 m 3 ; 1.6 × 3 × 1.6 m 3 ) to simulate pedestrians, vehicles, and other obstacles on the road. During the driving process of the vehicle (1 m–50 m), the average SE of ranging results in the channels is 0.72 m and the average RMSE is 0.77 m. After data correction, the average SE is reduced to −0.06 m, and the average RMSE is reduced to 0.11 m. The extreme difference of SE is reduced from 0.279 m to 0.025 m, and the decrease rate is 90.9 %. The extreme difference of RMSE is reduced from 0.166 m to 0.015 m, and the decrease rate is 91 %. In addition, we verified the applicability of the algorithm in four environments consisting of sunny day, sunny night, rainy day and rainy night. In these environments, this method still reduced the differences between channels and improved the ranging accuracy.
optics,physics, applied