A Method for Estimating Normal Vectors of 3D Point Clouds

Abstract

The invention discloses an estimating method for normal vectors of points of three-dimensional point clouds. The estimating method comprises the following steps: determining a k nearest neighbor point of each point in a space neighbor field; calculating an initial normal vector of each point according to the k nearest neighbor point; fitting a plane; choosing the k nearest neighbor point with 60%-80% of a nearest distance according to a distance from the k nearest neighbor point to the plane; rechoosing the k nearest neighbor point forming 60%-80% of a smallest included angle with the normal vector of the k nearest neighbor point chosen for the first time; eliminating noise points; calculating the stable normal vectors of the chosen points; and traversing the point clouds so as to obtain the stable normal vectors of each point. The estimating method successively uses the distance and the direction consistency to judge and eliminate unstable points in a K nearest neighbor, so that the more stable normal vectors of the points in the point clouds can be obtained by calculating, the more stable calculating parameters can be provided for the registration, clustering, division and the like of the point clouds, and the inhibiting ability for noises can be improved.

Description

technical field [0001] The invention relates to a parameter calculation method, in particular to a three-dimensional point cloud point normal vector estimation method. Background technique [0002] With the advancement of sensor technology, the acquisition of 3D point cloud is becoming easier and easier. How to understand the 3D scene represented by point cloud is an important problem to be solved in the fields of intelligent robot navigation, environment modeling, and somatosensory games. It is difficult to calculate and analyze a point cloud scene simply by using the three-dimensional space coordinates of points. Using geometric reasoning technology to extract compact and effective point features can better describe the point cloud and facilitate the next step of calculation. The normal vector of a point is the most widely used feature and is widely used in data registration, segmentation, etc. [0003] At present, the method of calculating the eigenvalues ​​and eigenvect...

AI Technical Summary

Problems solved by technology

[0003] At present, the method of calculating the eigenvalues ​​and eigenvectors of the covariance matrix of a point and its neighbors is commonly used to estimate the direction of the normal vector of the point cloud. This method is sensitive to noise, and the point cloud data itself often contains a lot of noise, so Using this method is easy to introduce a large error when estimating the normal vector of the point cloud.

Images