Research On Mapping And Feature Matching Method Of Single-Line Lidar

The development of logistics intelligence and unmanned technology has raised higher demands for the intelligence of robots.Localization and navigation,as a key technology for robot intelligence,have become a research hotspot.Single-line Lidar,due to its advantages of high accuracy,high resolution,wide measurement range,and good environmental adaptability,has been widely used in the field of localization and navigation.The research on single-line Lidar localization is one of the important research contents in the field of localization and navigation,which mainly includes mapping and matching.In order to improve the positioning accuracy of single-line Lidar,this paper conducts research on key technologies of feature mapping and feature matching,with the following main contents:1.Improvement of feature extraction and feature mapping algorithms for single-line Lidar.Based on the analysis of the limitations of traditional slope difference feature extraction algorithms,an improved feature extraction algorithm is proposed.This algorithm first uses the slope difference method to preliminarily determine the corner points and breakpoints,then uses the continuous edge tracking method to determine the final breakpoints,and uses the Iterative End Point Fit(IEPF)algorithm to determine the final corner points,and finally fits the feature lines using the least squares method.Feature extraction experiments have been conducted to verify the accuracy and stability of the improved feature extraction algorithm in recognition.Based on this feature extraction improvement algorithm,a feature map is established.2.Improvement of single-line Lidar feature matching algorithm.To address the limitations of the complete line matching method,a matching method based on a bounding box is proposed,and an improved Lidar line segment feature matching method based on Kalman fusion is proposed on top of this method.First,the improved Lidar line segment feature extraction method is used to extract feature segments and obtain a local map.Then,the local map is matched with the global map to obtain the pose change observed by the Lidar,and the pose change is predicted using IMU data.Finally,Kalman fusion is used to predict the optimal estimated pose from the predicted and observed poses.3.Real vehicle experiments were conducted to verify the above algorithm.An AGV indoor navigation system was built for testing the single-line Lidar feature mapping and matching algorithm.An experimental plan was designed,and positioning data of the AGV car under four different working conditions,namely X,Y,K directions and circular motion,were collected.The results of the data comparison and analysis under the four working conditions showed that the improved algorithm had a more significant improvement in both trajectory smoothness and positioning deviation compared to the slope difference algorithm.Conclusion: In response to the needs of single-line Lidar feature mapping and matching,this paper proposes an improved feature recognition algorithm and an improved Kalman fusion-based feature matching algorithm.The improved feature recognition algorithm improves accuracy and stability.The improved Kalman fusion-based feature matching algorithm improves matching precision.The experimental results verify the effectiveness of the improved algorithms,with a maximum error of 5.5cm for the improved algorithm and 30 cm for the slope difference algorithm.The average error of the improved algorithm can be controlled to around 1cm,while the fluctuation range of the slope difference algorithm is 1-8cm.