Indoor 3D Modeling Method Based On SLAM Laser Point Cloud

The three-dimensional model of indoor environment is widely used in smart city,indoor navigation and positioning,emergency response,etc.It can provide important technical support for location-based services and complex spatial analysis.It is important to study fast and efficient indoor 3D model construction methods.In this paper,the indoor point cloud data acquired by Simultaneous Localization and Mapping technology based on the backpack-type 3D laser scanner.The research improves the point cloud denoising and smoothing method based on Statistical Outlier Removal and Moving Least Squares,the indoor plane segmentation algorithm based on region growth,and the boundary extraction methods.In this way,a method for constructing a three-dimensional wireframe model of an indoor scene is established.The main research contents of the thesis are as follows:(1)Research the 3D laser point cloud mapping based on Lidar Odometry And Mapping.Most of the traditional point cloud registration methods have problems of high initial position requirements and complicated technical methods.In this paper,SLAM technology is introduced into the point cloud mapping of indoor three-dimensional scene.This paper analyses the basic principle of LOAM algorithm,completes the registration of indoor 3D point cloud using this algorithm,and evaluates the accuracy of registration point cloud data.The registration results are analyzed by comparing the measured results with the experimental results.The experimental results show that LOAM algorithm can achieve better point cloud mapping and obtain high quality three-dimensional point cloud data.(2)Research on point cloud denoising and smoothing based on SOR and moving least squares.Aiming at the problems of outlier noise and double-wall artifacts in SLAM point cloud data,the basic principles of SOR algorithm and MLS algorithm are compared and studied.The combination of SOR algorithm and MLS algorithm is used to denoise and smooth SLAM point cloud.Firstly,based on the statistical analysis method,the number of neighborhood points(k)and the standard deviation multiple(n)are set to remove the outlier noise points with large errors in the point cloud.The experiment shows that k = 30,n = 2 has the best denoising effect.Then point cloud smoothing based on MLS algorithm is used to further remove the small irregular noise in point cloud data and smooth the point cloud.The experimental results show that the method can effectively remove point cloud noise and reduce the surface thickness of point cloud.(3)Reconstruction of 3D wireframe model for indoor environment.SLAM point clouds are segmented based on region growth algorithm,and then the segmented point clouds are classified into top,ground,wall,door and window,etc.The boundary points of walls,doors and windows are determined,and the boundary points are processed by linear fitting.At the same time,the plane fitting of the segmented wall surface point cloud is carried out,and the common intersection point and intersection line of the wall are determined by fitting the relationship between the plane.Finally,the three-dimensional wire-frame model of the indoor scene is constructed.The model is compared with the actual measurement model,and the analysis shows that the median error of extracting wire frame model is 0.015 m.Through further manual editing,the wireframe model can meet the needs of various application scenarios to a certain extent.