Research And Implementation Of The Indoor Mobile 3D Mapping Based On Laser And Vision

With the development of science and technology,compared with 2D(two-dimensional)plane data,the 3D(three-dimensional)data is more real,has more information and so on.It is being widely used in various fields.In the indoor environment,how to locate the data acquisition platform accurately and complete high-precision data acquisition at the same time are the keys to the production of 3D map.Combined with high-precision LiDAR(Light Detection and Ranging),the technology of SLAM(simultaneous localization and mapping)is becoming an important development direction in the field of indoor 3D mapping.Based on the 2D laser SLAM,this paper studies the use of the panoramic camera(Ladybug5)and two two-dimensional LiDAR(HOKUYO UTM-30LX)in the indoor environment without GPS,inertial navigation system to complete the relative high precision 3D map production.The main research contents are as follows.(1)According to the project requirements to establish a new indoor high-precision three-dimensional map production program,which includes data acquisition,data processing,data display and data accuracy verification.Data acquisition uses two vertical 2D LiDAR to collect 3D laser data and uses panoramic camera to collect the panoramic images.Firstly,this paper introduces how to use the laser SLAM based on Rao-Blackwellized Particle Filter to create a grid sub-maps in the unified coordinate system.Through the data acquisition in the indoor environment to generate the 2D map,the correctness of the 2D-SLAM scheme is verified.(2)In the two-dimensional map optimization process,on the one hand,through the reverse thinking of based on landmarks SLAM,this paper proposes the use of virtual landmarks,there is no need to set the landmarks in the real scene and add optimization constraints.Combined with the graph optimization theory,the accuracy of the map is improved by minimizing the residual equation,On the other hand,inspired by the visual SLAM,this paper proposes the use of visual closed-loop to assist laser data for closed-loop determination.Through the closed-loop detection,it can eliminate the cumulative error as much as possible to the two-dimensional composition.(3)In the process of data fusion,by using RANSAC algorithm to obtain the laser spot with the weight to reduce the influence of the noise and the deformation of the calibration plate,the accurate extrinsic parameters among the three sensors are obtained by improving the optimization function,this paper uses extrinsic parameters for laser and image data fusion to get three-dimensional map.(4)The data display is divided into two-dimensional map and three-dimensional map data display,respectively,through the software platform and web platform.The data display platform has the function of measuring distance,which makes the point of the map can be measured directly to get the measured value.The Electronic Total Station is used to verify the data accuracy,corresponding to the points which have been chosen in the display platform,the laser targets are placed at their place to measure the distance between them as the true value.The relative accuracy of the map is verified by comparing the measured value with the true value.