关键词: 外廓尺寸, 激光雷达, kd-tree, 点云数据, 最小二乘, 三维轮廓重构

Abstract: A dynamic measurement method based on laser point cloud was proposed in order to solve the problem of low measurement accuracy，poor repeatability and poor reconstruction quality of three-dimensional ( 3D) contour in the measurement of vehicle contour dimension． Firstly，the method was based on the measuring principle of vehicle contour dimension，and the system measurement scheme was designed． Then，in order to remove the noise and redundant data generated by lidar during data collection，the topology of the point cloud was established based on kd-tree and the neighborhood averaging method was used to denoise the point cloud data，and the point cloud data were simplified by using the least square method to judge the local curvature features，and the boundary features were protected by the boundary point recognition algorithm． Finally，the proposed method was verified by the real vehicle test，and the reflector filter and the curve correction model were designed to further optimize the test result． The test results show that the indicating errors of four types of vehicles are all less than 1% ，with a maximum repeatability of 0. 48% ，which has higher accuracy and better stability to meet the national standard requirements． According to the 3D contour reconstruction model of vehicle，it can locate out-of-gauge position quickly．

Key words: contour dimension, lidar, kd-tree, point cloud data, least square, 3D reconstruction