Workpiece three-dimensional point cloud registration method based on direction histogram signature features

Abstract

The invention discloses a workpiece three-dimensional point cloud registration method based on direction histogram signature features, and belongs to the field of machine vision. The method comprisesthe steps: extracting the key points of a workpiece target point cloud through an internal signature description algorithm; adopting a direction histogram signature feature descriptor to carry out feature description extraction on the key points; randomly sampling the key points, and searching and extracting registration point pairs by combining secondary constraints, namely neighborhood constraints and distance constraints, in offline calculated template feature description; solving a rotation matrix and a translation vector by utilizing dual quaternion to realize initial registration of thepoint cloud, and realizing accurate registration of the point cloud by utilizing an iterative closest point algorithm on the basis of an initial pose provided by the initial registration. Compared with conventional registration based on a two-dimensional image, the method has the advantages that the influence of illumination can be avoided, the requirement for the placement position of an object is lowered, and the accurate pose relation between the template and the target, namely the rotation matrix and the translation matrix between the template and the target, can be obtained.

Description

technical field [0001] The invention belongs to the field of machine vision and relates to a three-dimensional point cloud registration method of workpieces based on signature features of direction histograms. Background technique [0002] In recent years, machine vision technology has been gradually applied in industrial production, and the positioning and grasping of three-dimensional target workpieces by manipulators through vision technology is a current research hotspot. The key step in positioning and grasping is to obtain the pose relationship between the workpiece template and the workpiece target, that is, the rotation matrix and translation matrix between the two. Most of the traditional solutions use registration on two-dimensional images to obtain Pose relationship, this method is very sensitive to lighting and object placement, and positioning is prone to deviations. [0003] With the maturity of 3D technology, it becomes possible to solve the registration prob...

AI Technical Summary

Problems solved by technology

The key step in positioning and grasping is to obtain the pose relationship between the workpiece template and the workpiece target, that is, the rotation matrix and translation matrix between the two. Most of the traditional solutions use registration on two-dimensional images to obtain Pose relationship, this method is very sensitive to lighting and object placement, and positioning is prone to deviations

Images