A 3D Object Recognition Method Based on Binary Descriptor

Abstract

The present invention proposes a method for recognizing a three-dimensional object based on a binary descriptor, which includes a three-dimensional object feature description step, and the three-dimensional object feature description step includes the following process: S1, meshing the grid depth image Q in the three-dimensional object Calculation of resolution R; S2, uniformly sample the grid depth image Q according to the calculated grid resolution R, and obtain the coordinates of the feature point P'; S3, construct a local reference according to the calculated coordinates of the feature point P' system, and disambiguate the directions of the x-axis, y-axis, and z-axis of the local reference system; S4, construct a feature vector according to the disambiguated local reference system, and obtain the binarized three-dimensional feature fp of the feature point P' . The invention has simple calculation method, high efficiency of feature extraction and feature matching, and can quickly and accurately realize the binary description of the three-dimensional object.

Description

technical field [0001] The invention relates to the field of machine vision, in particular to a three-dimensional object recognition method based on a binary descriptor. Background technique [0002] The goal of 3D object recognition is to correctly identify the object and determine the position and attitude of the object in the 3D scene. Compared with the target recognition technology based on two-dimensional images, 3D object recognition has its unique technical advantages: first, it contains more depth information; second, feature extraction is less affected by scale, rotation and illumination; The estimation of target pose is more accurate. It is expected to solve many problems faced in target recognition based on two-dimensional images. In addition, the substantial improvement in computer performance and the reduction in the cost of depth data sensors in the past two decades have made 3D object recognition a major research hotspot in the field of machine vision. [0...

AI Technical Summary

Problems solved by technology

[0004] However, existing feature description methods usually use a row of floating-point vectors to represent the features of 3D local surfaces, and the feature vector dimensions are generally high

Therefore, the amount of calculation for feature matching is large, which requires high computing performance of the equipment, and the application scenarios of the algorithm are limited; the recognition process takes a long time, which is not conducive to the real-time realization of 3D target recognition

Images