| With the development of robot technology,the research of outdoor 3D Simultaneous Localization and Mapping using LIDAR has become a research hotspot.Compared with the indoor environment,outdoor environment is more complex,which requires more precise of technologies for its localization and mapping.Moreover,3D map with high precision can well characterize the outdoor environment,making outdoor path planning and autonomous navigation possible.In comparison to other sensors,3D LIDAR has many advantages,such as wide field of vision,high ranging accuracy,insusceptible to sunlight and strong anti-interference,which is especially suitable for SLAM in outdoor environment.Therefore,this study intends to investigate outdoor 3D SLAM including point cloud pre-processing,point cloud registration,loop closure detection and global optimization.The main research contributions are as follows:(1)LIDAR point cloud data pre-processing.Firstly,the hardware and software platform of point cloud data processing are introduced.Then,statistical filter is employed to filter outliers generated in the scanning process.In addition,voxel grid filter is applied for downsampling the point cloud to reduce the amount of point cloud data.(2)The point cloud registration algorithm based on LIDAR.First,the basic principles of common point cloud registration algorithms are studied.Next,the registration accuracy and timeliness are determined in experiments.The results show that the performance of feature-based point cloud registration algorithm is more excellent.Then,LIDAR odometry based on feature matching is proposed.Specifically,after the ground point cloud is quickly acquired by projection,the extraction of edge and plane feature points is completed based on the point roughness.The feature matching of adjacent point clouds is achieved by point-to-line and point-to-plane constraints.Furthermore,the 6-DOF pose is iteratively optimized by fractional steps,which improves the operation efficiency and accuracy.Furthermore,after comparing LOAM with the proposed method,it turns out the letter is more precise and efficient.(3)The loop closure detection algorithm and global optimization based on intensity scan context.In order to solve the problem of increasing cumulative errors in long running,a loop closure detection algorithm based on intensity scan context is proposed.In order to improve the efficiency and accuracy of loop closure detection,the candidate loop key frame is first determined by neighborhood search,and then the occurrence of loop return is confirmed according to the geometric characteristics of strength,followed by adding corresponding loop closure constraints.Finally,the globally consistent pose and map are obtained by global optimization.(4)The KITTI data set and real campus data are tested to verify the performance of the proposed approach.The obtained data from experiments indicate that the proposed method has higher localization accuracy and can construct high-precision maps in real time. |
