Visual repositioning method, system and device based on scene coordinate regression network

Abstract

The invention belongs to the technical field of service robots, particularly relates to a visual repositioning method, system and device based on a scene coordinate regression network, and aims to solve the problem that the visual repositioning method based on the scene coordinate regression network is difficult to adapt to large camera visual angle changes. The method comprises the following steps: acquiring an original color image of the surrounding environment of the service robot; constructing a first coordinate graph, a second coordinate graph and a distance graph; obtaining a source feature map through feature extraction; performing feature transformation on the source feature map; adding the transformed feature map and the source feature map, and performing regression through a regression device to obtain a 3D coordinate map and an uncertainty map; obtaining a 2D pixel coordinate graph, and filtering the 3D coordinate graph and the 2D pixel coordinate graph based on the uncertainty graph; acquiring the pose of the service robot through a PnP algorithm so as to achieve the visual repositioning of the service robot. According to the invention, the robustness of the visual repositioning method based on the scene coordinate regression network to large camera visual angle changes is improved.

Description

technical field [0001] The invention belongs to the technical field of service robots, and in particular relates to a visual relocation method, system and device based on a scene coordinate regression network. Background technique [0002] In recent years, with the rapid development of artificial intelligence, sensing technology, computing processing and other technologies, service robots are also developing in the direction of intelligence. Among them, robot positioning is the prerequisite for completing tasks, which requires the robot to use its own sensors ( For example, a visual sensor (i.e. a camera) determines its current position. For vision-based robot positioning, during the movement of the robot, sometimes the positioning fails. At this time, the robot can be repositioned according to the pre-built environment map and the image information acquired by the camera at the current moment (ie visual repositioning). In order to ensure the quality of robot task executio...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problem that the existing visual relocation method based on the scene coordinate regression network is difficult to adapt to large camera angle changes, the present invention proposes a visual relocation method based on the scene coordinate regression network , including the following steps:

Images