Ship pose estimation method based on three-dimensional point cloud features

Abstract

The invention provides a ship pose estimation method based on three-dimensional point cloud features, and the method comprises the steps: firstly obtaining the three-dimensional point cloud data of atarget ship from six directions, and obtaining a complete point cloud library of the target ship in a splicing mode; constructing a point cloud template library of the target ship; and calculating ISS3D feature points and feature histograms of the to-be-registered target ship point clouds to obtain an initial coordinate transformation matrix in which the feature points in the template point cloudlibrary coincide with corresponding feature points in the to-be-registered point cloud library, thereby achieving rapid coarse registration of the point clouds in a global range. adopting an ICP pointcloud registration algorithm to carry out accurate registration of the point cloud on the basis of rough registration in the last step, and obtaining a coordinate transformation matrix for achievingaccurate registration of the point cloud; and finally, performing six-degree-of-freedom pose estimation on the target ship by utilizing the precise coordinate transformation matrix and the initial pose information of the ship in the point cloud template library. The invention provides a ship pose estimation algorithm experiment based on point cloud features, and the effectiveness of the method isverified through experimental data comparison.

Description

technical field [0001] The invention relates to a ship pose estimation method based on three-dimensional point cloud features, which belongs to the signal processing technology method. Background technique [0002] 3D lidar is a very important sensor for obtaining surrounding information. It has the advantages of high measurement accuracy, high point cloud density, fast speed, and low cost. important role. Especially for ship-borne lidar, the three-dimensional point cloud information of the target ship is obtained in open seas, narrow waterways, and in and out of the port, and on this basis, the position and attitude of the target ship can be accurately estimated using the point cloud information. If the pose estimation error of the point cloud of the target ship is large, it will not only lead to a large error in subsequent processing, but may even cause a ship collision accident. Therefore, it is necessary to accurately estimate the position and attitude of the target shi...

AI Technical Summary

Problems solved by technology

The above literature has the following problems in the process of 3D point cloud target pose estimation: First, the estimation method based on deep learning requires a large number of samples to train the network, and collecting a sufficient amount of ship training samples is difficult and expensive to operate; The second is that the method based on viewpoint features needs a feature database composed of viewpoint feature maps as the basis. When the viewpoint feature map of the ship to be tested in the feature database is incomplete, it cannot provide accurate pose estimation; the third is the traditional method based on viewpoint features. The point cloud registration process is computationally intensive and takes a long time to run

Images