Research On LiDAR Odometry Optimization And Mapping Technology Based On Loop Closure Detection

With the rapid development of science and technology,all kinds of mobile robot technologies,such as self-driving and service robots,are making continuous progress driven by science and technology.SLAM algorithm is the core technology to realize the autonomous movement of robot,and it has important research significance.The SLAM algorithm shows that the robot can complete the map construction function when the position of the robot is uncertain,and use the map information to achieve autonomous positioning and navigation tasks.Multi-line Li DAR is the most commonly used sensor in SLAM algorithm.In the case of lack of texture or serious lack of light,the localization and mapping effect of laser SLAM algorithm is better than that of visual SLAM algorithm.LOAM algorithm is the most representative algorithm in laser SLAM algorithm.However,this algorithm does not take into account the point cloud intensity and does not make full use of the point cloud information in historical frames,so its odometry and mapping accuracy is insufficient.Aiming at the above problems,this paper proposes a Li DAR odometry optimization and mapping algorithm based on loop detection.The main research contents are as follows:1、 Considering that directly using all the laser point cloud data obtained by Li DAR for Li DAR Odometry and back-end optimization will greatly affect the efficiency and positioning accuracy,so the preprocessing operation is performed on the original laser point cloud data,and the distant and The closer point cloud achieves down-sampling,and at the same time filters out small objects,avoiding unreliable laser point clouds from affecting the accuracy of subsequent registration.2、 Making full use of the ground point cloud information,the ground points are extracted separately and applied to the back-end constraints.because the object intensity information of different materials is different,the intensity value and geometric value of the point cloud are considered to complete the corner and plane point extraction.3、 Considering that LOAM algorithm mainly uses geometric features for Odometry and mapping optimization while ignoring the intensity information of Li DAR point cloud,it is proposed to use geometric information and point cloud intensity information to realize Li DAR Odometry and mapping algorithm.Geometric information is used to optimize Li DAR Odometry and mapping through corner and face constraints,and reflectivity map is constructed using intensity information.The measured value of the reflectivity of the corresponding grid element is obtained by trilinear interpolation to realize the point cloud intensity constraint,and the point cloud intensity information is used to optimize the Li DAR Odometry and mapping.4、 Add loop closure detection module,on the basis of Li DAR Odometry optimization and mapping algorithm based on geometric information and intensity information,make full use of historical frame point cloud information,through loop information to improve the pose accuracy of Li DAR Odometry and mapping algorithm again.The loop frame data is obtained from the surrounding N-frame point cloud,and the ICP score is used to determine whether to optimize it or not.If optimization is needed,the loop position is corrected by point cloud intensity constraint and ground plane constraint on the basis of the position obtained by ICP algorithm.Finally,the loop closure detection factor is added to the factor graph algorithm to complete the correction of the non-loop region data frame,and the global consistent map is obtained.5、 The accuracy of the improved algorithm is verified by using KITTI open dataset.In a variety of road scenarios,the results of IG-LOAM algorithm with geometric constraint,intensity constraint and ground constraint are compared with those of IGCF-LOAM algorithm with geometric constraint,intensity constraint,ground constraint,loop constraint and factor graph,and the results of ALOAM algorithm are compared.Absolute pose error and relative pose error are used to evaluate the effect of the algorithm.And with the help of ROS operating system tools to show the experimental results.Through the comparison of the data indicators in the error,it can be verified that the improved algorithm can greatly improve the effect of location and mapping.