Research On 3D Scanner System Based On Laser And Image Fusion

During the construction of a project,the existing building structure body needs to be mapped to obtain the three-dimensional coordinate points of the building.After the project is completed,the finished building is restored in 3D and the computer software simulates to test the reliability of the structure.When the large-scale structure is reconstructed in 3D,it will face the problem that the building area range is large,and the constructed 3D model has low accuracy and serious deformation,which resulting in the model cannot be utilized.With the continuous in-depth exploration of building restoration technology,researchers have found that the 3D building and positioning system can effectively solve the problems faced in the current engineering construction instead of the traditional 3D scanning technology.The purpose of 3D mapping and positioning technology applied to robot motion is to realize that a robot entering an unknown environment can use its own sensors(e.g.,depth camera or radar)to capture point cloud information in the external environment and obtain a map of the current environment after data pre-processing,inter-frame matching,back-end optimization,loopback detection,and map building by the host computer.The environment map includes the location,direction and information of the environment in which the robot is located in space.After the robot receives the movement command,it performs point-to-point,global or local path planning on the existing map to calculate the optimal collision-free path between the current position and the target point to realize robot positioning and navigation.So far,the introduction of robot map building and positioning systems in construction projects has encountered three problems: 1)the point cloud information obtained by a single 3D laser scanner does not contain object surface texture information,which is not conducive to data feature extraction;2)the traditional total station single point measurement mode obtains a large amount of point cloud data,which is difficult to process;3)when measuring large buildings,the size error of the building due to the complex structure large and not accurate enough to complete matching.To address the above problems,this thesis first analyzes the current market common 3D scanner products,summarizes the problem points with the actual project,designs the scanning system according to the laser map building and positioning technology,and builds the mobile robot experimental platform.Then,the Kinect depth camera image is calibrated based on the hand-eye calibration method to achieve accurate scanning of environmental objects,and GPS is used to fuse color point cloud information of the same scene with 3D laser point cloud information to enhance point cloud visualization,contain more feature information,facilitate inter-frame matching and provide higher precision alignment initial values.Finally,the point cloud fragment feature verification method is used as a loopback detection module added to the LOAM algorithm to analyze the track error of LOAM,this paper system and standard GPS track error,the track drift error rate is reduced from 0.71% to 0.42%,the root mean square error and the mean value are reduced by nearly half,the track deviation occurred during map construction is reduced,and the accuracy of map construction is significantly improved.The octree map saves point cloud data to reduce computer storage space by 6 times.The map building system of this thesis gives a new solution to the study of construction engineering measurement and provides a 3D topographic digital reconstruction method to further promote the construction of modern urban digital development.