Visual navigation method based on fast feature point extraction and mesh triangle constraint

Abstract

The invention discloses a visual navigation method based on fast feature point extraction and mesh triangle constraint, which comprises the following steps: acquiring continuous video frame sequence images, extracting feature points in adjacent two frame images; meshing the extracted two adjacent frames of images; a 6-dimensional shape descriptor and a 32-dimensional region descriptor being constructed for each triangle, and a feature vector composed of 38-dimensional mixed descriptors of each triangle being obtained; matching the eigenvectors of the triangles corresponding to the mesh division, selecting the feature points with remarkable characteristics by the center point clustering method, eliminating the mismatched feature points, and eliminating the motion feature points by the polargeometry constraint, so as to obtain the effective and usable feature point pairs; completing the final solution of the motion model, and obtaining the position results of visual navigation solution.The invention provides a solution for robot vision navigation and positioning under indoor environment, and has the advantages of fast feature point extraction speed and high matching rate.

Description

technical field [0001] The invention relates to the technical field of visual navigation and image processing, in particular to a visual navigation method based on fast feature point extraction and gridded triangle constraints. Background technique [0002] In the field of robot autonomous navigation and positioning, the most widely used method is the navigation method relying on GPS and inertial system. With the rapid development of science and technology, many emerging navigation methods have gradually emerged, such as visual navigation. In recent years, with the rapid development of machine vision technology, more and more robots have begun to use machine vision for autonomous obstacle avoidance and path planning. According to previous navigation methods, most robots use GPS for positioning and navigation. However, in the field of indoor navigation, it is difficult for GPS signals to penetrate thick walls to transmit signals for robots, which will seriously affect the re...

AI Technical Summary

Problems solved by technology

[0003] The extraction and matching of image feature points in visual navigation is a very important part of the implementation, and image feature points are generally detected at multiple scales, so many points with very close positions and scales will be caused, resulting in the phenomenon of feature point aggregation, forming redundant Additional feature points, increasing the probability of false matching

Images