Research And Implementation Of Key Techniques Of 3D Point Cloud Surface Reconstruction On Vehicle Dimensions Measuring System

As the productions and usages of vehicles are difficult to get standardized controls,the phenomena of illegally spliced and assembled vehicles often occurs and results in problems with overlength,overwidth and overheight vehicles.Overrun transportation is the main factor causing overload transportation.Overrun and overload transportation can cause traffic accidents and damage road facilities.At present,when measuring vehicles' overall dimensions,the vehicle administrative offices and vehicle detection institutions in China mostly use manual measurement methods,sometimes they even rely on visual experience.These methods have the disadvantages of large error,low efficiency,high labor intensity and high miss rate,and they can not measure moving vehicles.To solve the problems above,this thesis presents a scheme for vehicle profile dimensions measurement based on laser scanning technique,and completes the development on the prototype of the vehicle dimension automation measuring system.After real vehicle measurement experiments,we verified the effectiveness and stability of system's functions.The system measuring error is within ±1%when the car speed is below 20km/h,it contents the demands on vehicle dimension automation measuring systems claimed in the national standard "GB21861-2014".The system provides effective methods of the management on vehicle modification and overrun for the departments of road transportation and traffic management.The main work and achievements of this thesis are as follows:1.This thesis analyses and researches the methods for vehicle profile dimensions measuring both at home and abroad and the operational principle of LMS111 laser ranging sensor.The scheme of vehicle dimension measuring system is developed.This system uses two laser ranging sensor LMS111 and one radar tachometer to measure the width,height and length of vehicles.2.This thesis researches key technologies and principle of main algorithms of laser scanning point cloud data process.Based on PCL,a open source C++ programming library,the point cloud data preprocessing program is designed and implemented,including point cloud filtering,point cloud segmentation and point cloud registration.The extraction of effective vehicle data from massive collected point cloud data is achieved,it establishes the basis of vehicle profile surface reconstruction.3.This thesis intensively researches the principle of point cloud surface reconstruction,and implements the design of vehicle profile surface construction program module based on the greedy projection triangulation algorithm.This thesis completed point cloud smoothing program based on least square method,aimed at the problem that the redundances of point cloud influence the smoothness of reconstructed model.Finally this thesis completes the implementation of the effect-optimized vehicle point cloud surface reconstruction program.The result of surface construction on real vehicle data shows that the vehicle profile surface reconstruction program has a actual and beautiful effect on vehicle profile surface construction.4.The design of vehicle dimensions database is implemented based on the Qt development platform.It supports the storage,add,query or delete operations on real vehicle measurement data.5.This thesis finished the experimental scheme of real vehicle measurement,analyses the source of error through measuring data,revised system software code and improves the experimental scheme.The system response time and error are reduced,and the accuracy of measurement is increased.The experimental results indicate that the performances of the vehicle dimensions measuring system in this thesis satisfies the design requirements and the measurement accuracy meets the demands of national standards.This thesis realizes the goal of the R&D design.