A mobile robot positioning method based on an improved ORB algorithm

Abstract

The invention discloses a mobile robot positioning method based on an improved ORB algorithm and belongs to the technical field of positioning technology of mobile robots. and the method comprises a mobile robot and a computer, and steps of inputting of image pairs: the mobile robot acquires images in real time through a camera and transmits the images to the computer for image processing in realtime; Image gridding: using a GMS algorithm to segment the image into non-overlapping units with G = 20 * 20 by using a grid; ORB descriptor: detecting the extracted feature points by using an ORB algorithm; According to the invention, feature points with scale invariance can be extracted by using an SIFT algorithm; and finally, the matching quality is remarkably improved by utilizing a GMS feature registration algorithm based on a scoring framework, the real-time performance and the feature matching accuracy are better, the odometer accumulation error in autonomous navigation is effectively reduced, and the autonomous positioning precision of the mobile robot is improved.

Description

technical field [0001] The invention relates to the technical field of mobile robot positioning, and more specifically, relates to a mobile robot positioning method based on an improved ORB algorithm. Background technique [0002] In recent years, with the development of artificial intelligence technology, mobile robots have been widely used. It liberates people from heavy physical labor and effectively saves labor costs. Mobile robots need to obtain information in real time during their long-distance travel. Positioning information, commonly used mobile robot navigation methods include magnetic navigation, inertial navigation, visual navigation, etc. Magnetic navigation has high reliability, but it is costly and difficult to maintain; inertial navigation is a common navigation method for mobile robots, because it will lead to accumulation The error increases infinitely and is not suitable for long-term precise navigation. Currently, the commonly used mobile robot navigation...

AI Technical Summary

Problems solved by technology

[0002] In recent years, with the development of artificial intelligence technology, mobile robots have been widely used. It liberates people from heavy physical labor and effectively saves labor costs. Mobile robots need to obtain information in real time during their long-distance travel. Positioning information, commonly used mobile robot navigation methods include magnetic navigation, inertial navigation, visual navigation, etc. Magnetic navigation has high reliability, but it is costly and difficult to maintain; inertial navigation is a common navigation method for mobile robots, because it will lead to accumulation The error increases infinitely and is not suitable for long-term precise navigation. Currently, the commonly used mobile robot navigation method is a grid-based motion statistics combined navigation method. Although this navigation method has high stability, it has a large cumulative error

[0003] The ORB algorithm is a feature point detection and description algorithm based on visual information. It uses the FAST (features from accelerated segment test) algorithm to detect a certain feature point and the pixel value in the surrounding area. The area is generally a circular area. , the ORB algorithm uses the BRIEF (binary robust independent elementary features) algorithm for feature point description, selects several pairs of random points near the feature points, compares the pixel size of each pair of random points and performs binary assignment, and then combines them into a binary code, which is The feature descriptor sets the threshold according to the Hamming distance between the matching point and the non-matching point to achieve matching. The ORB algorithm adds direction information to the FAST feature. At the same time, the BRIEF descriptor is used in the feature description and is improved. Its descriptor has better robustness to noise and rotation. Although the ORB algorithm has a faster calculation speed and has better performance on noise and image rotation changes, but because the FAST feature does not have scale invariance, so Not robust to image scale changes

[0004] Positioning based on visual odometry is a relatively new mobile robot positioning technology at present. The robot collects image features through the camera, performs pose estimation and autonomous navigation according to the fusion information of image features and motion constraints, and random sample consensus (RANSAC) ) algorithm is one of the commonly used visual feature registration algorithms, but the RANSAC algorithm has a high error matching rate and complexity

Images