Positioning method and system

Abstract

The invention discloses a positioning method and system. The method includes at least the following steps: acquiring an image sequence; extracting 2D feature points and from images of adjacent two frames in the image sequence and performing matching; acquiring 3D feature points of the extracted and matched 2D feature points; on the basis of inter-frame motion displacement measurement and the number of internal point that are selected, matching the 2D feature points which correspond to the 3D feature points with the features points of a current frame by means of the optical flow method or the feature matching method, obtaining a rotation matrix of a current moment and a transverse vector of the current moment; on the basis of the rotation matrix of the current moment and the transverse vector of the current moment, obtaining a positioning position. According to the invention, the method can selects the computing result of a corresponding algorithm in a flexible and alternative manner by separately resolving the inter-frame motion displacement measurement and the number of internal points that are applied by the two methods. Therefore, the method and the system accelerate positioning, increases positioning precision, and has strong practicality in complex environment.

Description

technical field [0001] The invention relates to the technical fields of image processing and computer vision, in particular to a positioning method and system. Background technique [0002] Visual odometry only obtains the visual information obtained by the moving object or robot camera to complete the positioning of the moving object or robot. The disadvantages of traditional odometers are poor measurement accuracy and unstable data, while visual odometry can effectively overcome the problems of wheel slippage and long-distance errors of traditional odometers, and can solve the problem of precise and autonomous positioning of robots. This technology has been widely used in the positioning of ground robots, underwater robots, and micro-UAVs. It can be divided into binocular stereo vision odometers and monocular vision odometers. It was first proposed by Nister et al. in 2004. [0003] At present, the working methods of monocular visual odometry can be divided into three cat...

AI Technical Summary

Problems solved by technology

This is because, compared with 2D-2D, the 3D-2D method not only requires triangulation of points, but also has a relatively complex calculation method, and also needs to obtain an accurate 3D feature set and consider the noise problem of camera extrinsic parameters.

Images