Research On Lidar Imaging Simulation And Point Cloud Ship Position Estimation

Waterway transportation occupies an important position in international cargo transportation by virtue of its low cost and large cargo capacity.With the increase in international trade volume and the increasingly busy shipping routes,the inherent problems of high operating costs and high safety risks are becoming increasingly prominent.The development of unmanned intelligent ships with low pollution,high safety and high efficiency can effectively solve the problems above.Under this background,this paper carried out the research of using 3D lidar as the sensor to obtain the marine ship point cloud to estimate the ship’s pose and attitude.By building simulation software to obtain the point cloud data,the problem of insufficient data of the point cloud in the marine ship for research can be solved.The specific research content is as follows:First,the problem of insufficient point cloud data for ships on the sea is solved by software simulation.By writing a 3D simulation software to simulate the ship scene on the sea,the distribution of the ship is set up according to the requirements and displayed visually.Then by writing a program by simulating real radar parameters,the radar point cloud data under the current distribution can be captured.Through this software,enough point cloud data of the ship distribution scene can be get,and can be used as the research basis for point cloud data related algorithms.Second,research on point cloud segmentation algorithms.Preprocess the original point cloud using pass-through filters,voxel filters and radius filters,and use kd-tree to improve the speed of point cloud processing.Then,improve the segmentation method based on regional growth by referring to the idea of Euclidean clustering.By dynamic adjustment of the seed point growth region,the problem of easy over-segmentation of the traditional regional growth method is solved,the point cloud segmentation effect is improved,and the speed of point cloud segmentation is ensured.Finally,estimate the pose and attitude of the ship using the method based on point cloud registration and the method based on PCA,and analyze and summarize the two methods through comparative experiments.In the ship pose estimation method based on point cloud registration,the registration method combining SAC-IA coarse registration and ICP fine registration is used to estimate the ship`s pose.The SAC-IA matching efficiency is improved by the corresponding point pre-judgment method.The index speed of the point cloud during the ICP registration processis improved by introducing the kd-tree structure,and the registration effect is improved by adding the weight of the corresponding point clouds.This pose estimation method has high accuracy,but it has the disadvantages of large amount of calculation and slow calculation speed.Meanwhile,it needs to build a point cloud template library in advance.In the method of estimating the ship’s pose and attitude based on the PCA method,the rectangular feature of the ship is mainly used to calculate the main direction of the ship and the OBB bounding box.Then,the same ship in the multi-frame point cloud is matched using the PCA feature and the bounding box feature.Finally the ship’s navigation information can be calculated.The accuracy of this pose estimation method is relatively low,but the calculation speed is fast,and there is no need to build a point cloud template library in advance.