A method of constructing k-d tree based on point cloud registration icp algorithm

Abstract

The method for constructing a K-D tree based on the point cloud registration ICP algorithm provided by the present invention divides the point cloud space into multiple subspaces, defines the subspaces with data points as effective subspaces, and establishes K-D for the effective subspaces. D tree, in order to reduce the complexity of point cloud K-D tree, speed up the speed of hardware construction of K-D tree, and reduce the storage burden of hardware. This method uses direct storage to store point cloud data, and directly obtains the addresses of all data in the subspace according to the subspace coordinates, so as to use the advantages of hardware parallel processing to speed up the nearest neighbor query. This method is the hardware for accelerating iterative nearest point algorithm The system provides a solution.

Description

technical field [0001] The invention relates to the field of integrated circuit design, in particular to a method for constructing a K-D tree based on a point cloud registration ICP algorithm. Background technique [0002] A point cloud is a collection of a series of points in three-dimensional space, and the point cloud data representing the surface characteristics of the object can be obtained by scanning the surface of an object with a specific device. Any point in the point cloud is usually represented by coordinates in a three-dimensional rectangular coordinate system. The process of restoring the 3D model of an object based on the recognition, filtering, and registration of point cloud data is called 3D reconstruction. Point cloud registration is a critical step that affects the performance of 3D reconstruction. [0003] Point cloud registration is a technique to stitch two point clouds into the same coordinate system, which is divided into two steps: coarse registra...

AI Technical Summary

Problems solved by technology

For a large number of point clouds, this process takes a very long time and cannot be accelerated by taking advantage of hardware parallelism;

[0006] 2. In the prior art, the K-D tree nodes are stored in the form of a linked list. In addition to storing the data of the node, each node also stores the addresses of the parent and child nodes of the node. This indirect storage method not only increases the burden of hardware storage , which also greatly limits the speed at which the hardware can read data

Images