Research And Implementation Of Vehicle External 3D Feature Detection System Based On Dynamic Motion Information

Overloading of vehicles will not only damage road and bridge facilities,but also easily cause traffic accidents and even cause casualties,which has a huge impact on the development of national economy and social security.The existing over-the-limit detection system has uneven data acquisition speed,low detection accuracy,single detection content,and limited external three-dimensional feature parameters of the vehicle,which can only calculate the exterior dimension of the vehicle,and cannot quickly and accurately identify the external features of the vehicle.Aiming at the above problems,a set of vehicle external threedimensional feature detection system integrating dynamic motion information was designed and implemented,and the vehicle external dimension was accurately measured and the body three-dimensional reconstruction was realized.The main work completed was as follows:(1)Aiming at the problem that the detection accuracy is not high due to the mismatch between the acquisition speed of LIDAR and the moving speed of the vehicle,the Angle and acceleration information of the inertial measurement unit is integrated to correct the distortion of the original point cloud data collected,and the Li DAR is calibrated based on the calibration method of the Li DAR system of the cube.The measurement method of vehicle external dimensions is studied.After the initial screening and key frame extraction of the original point cloud data collected,the vehicle dimensions are measured by relevant calculation methods.In order to meet the synchronization requirements of the left and right lidar,the vehicle length,width and height detection method based on P-frame strategy is improved to detect the key information of vehicle profile dimensions.(2)A vehicle over-limit cargo detection system has been established,slice processing and contour extraction of vehicle point cloud data collected by the system;In order to solve the problem of low extraction efficiency of the Alpha-Shape algorithm when the density is inconsistent,an improved contour extraction algorithm based on adaptive values was proposed to detect the contour features of over-limit cargo.The improved K slice method is used to measure the volume of vehicles and over-limit goods accurately.(3)A three-stage body 3D feature reconstruction framework is designed.Based on the accurate measurement of the original vehicle point cloud,after background removal and redundancy frame elimination,the Filter Reg registration algorithm is applied for rough alignment,and the line process and G2 O map are combined to further complete the 3D matching optimization,so as to achieve the accurate reconstruction of the 3D feature point cloud of the large car body.(4)The actual vehicle experiment and test are completed,including point cloud correction experiment,vehicle size detection experiment and body reconstruction experiment.The experimental results show that the absolute error of vehicle profile detection can reach 5mm and the relative error can reach 1.4%,which improves the detection accuracy of vehicle external features,and the matching degree of reconstructed vehicle three-dimensional map can reach97.3%.