Matching method of feature points in robot vision navigation positioning image

Abstract

The invention provides a matching method of feature points in a robot vision navigation positioning image. The method comprises steps of extracting feature points from two frame images and performing rough matching of the feature points of the two frame images; and filtering the matching feature points using affine transformation invariance of spatial geometry. By using the feature that the captured ceiling image feature points have the affine transformation invariance between different frames, the mismatched feature points can be filtered according to whether the connection length of the feature pair in the two frame images and whether the corresponding angles are equal. The feature points on the ceiling can be screened, the accurate matching is realized, and the mismatched interference points can be removed. In addition, when the matching of the feature points of different frame images is made, that whether the length and the angle are equal are judged by using the threshold determination method, and the screening range of the effective feature points is enlarged, which is not limited to the ceiling feature points. The matching method is also applicable to the positioning in fuzzy, pixel discretization and other conditions of shooting.

Description

technical field [0001] The invention belongs to the field of machine vision positioning, and relates to a matching method for robot vision navigation and positioning image feature points. Background technique [0002] Mobile robot is a comprehensive system integrating environmental perception, dynamic decision-making and planning, behavior control and execution, etc. It is widely used in military, industrial and civilian fields. The positioning of mobile robots is an important basis for behavior control and execution, and environmental perception. At present, common positioning methods for robots are mainly divided into five categories: ultrasonic navigation and positioning, visual navigation and positioning, GPS global positioning, light emission navigation and positioning, and real-time positioning. And map construction (Simultaneous Localization and Mapping, SLAM). [0003] In the robot vision navigation and positioning technology, accurate autonomous positioning is the ...

AI Technical Summary

Problems solved by technology

For a robot moving in a building, the feature points on the ceiling are usually the most effective navigation feature points. As long as the feature points in the robot vision are matched in the map and the corresponding points of these points in the map are found, the It is convenient to calculate the position and heading of the robot on the map, but there are still the following technical problems to be solved urgently: the image collected is a two-dimensional image, lacking height information, and it is difficult to distinguish ceiling feature points from feature point information at other height positions; In addition, most of the current feature point extraction and matching algorithms are based on the color distribution in the neighborhood of the matching points, and there are many similar indoor scenes. Similar small areas cause great interference to feature point matching, which is easy to cause mismatching

Images