Visual odometer construction method in robot visual navigation

Abstract

The invention provides a visual odometer construction method in robot visual navigation. The visual odometer construction method comprises the following steps: extracting image feature points; obtaining the position of an object bounding box; judging whether the extracted feature points are in the detected object bounding box or not, and if not, removing the feature points; if the number of the dynamic points in the object bounding box is within the object bounding box, tracking the feature points in the object bounding box through optical flow, judging the dynamic points through an optical flow epipolar method, and if the number of the dynamic points in the object bounding box is greater than a threshold value, considering that the object in the object bounding box is moving, and removingall the feature points in the object bounding box. If the threshold value is not exceeded, it is considered that the object in the object bounding box is static, wherein the feature points in the object bounding box are effective feature points; estimating a camera motion track according to the effective feature points, recovering the rotation and translation motion of the camera, estimating thepose of the camera, and completing the construction of the visual odometer. According to the invention, the problem of inaccurate pose estimation in a dynamic environment is solved, and the positioning precision is improved.

Description

technical field [0001] The invention relates to the technical field of visual navigation and image processing, in particular to a method for constructing a visual odometer in robot visual navigation. Background technique [0002] After the robot enters the unknown environment, the positioning and navigation functions are inseparable from the visual odometer. The visual odometer refers to the processing and analysis of continuous video sequence frame images through machine vision technology to complete the position and attitude estimation of the mobile camera and realize the navigation and positioning functions. Traditional visual odometry has poor robustness and accuracy in dynamic environments. Moving objects in the scene will cause mis-matching of feature points in pose estimation, which in turn affects positioning accuracy and accuracy of mapping. It is affected by moving objects in the external environment. The interference of feature points is relatively large. Conten...

AI Technical Summary

Problems solved by technology

Traditional visual odometry has poor robustness and accuracy in dynamic environments. Moving objects in the scene will cause mis-matching of feature points in pose estimation, which in turn affects positioning accuracy and accuracy of mapping. It is affected by moving objects in the external environment. The interference of feature points is large

Images