Camera repositioning method based on semantic road sign

Abstract

The invention relates to the technical field of repositioning, and provides a camera repositioning method based on semantic road sign, which comprises the following steps: arranging semantic landmarks, and constructing a scene map; initializing a ground pre-selected area; performing real-time camera shooting on the scene by using a camera to form a scene image library, extracting angular points inthe scene image, estimating the current pose of the camera, and if the current pose can be estimated, constructing a three-dimensional point cloud of the scene; if the current pose cannot be estimated, the camera loses positioning, and repositioning the camera: extracting angular points in the current frame and the candidate frame; carrying out corner point matching on the current frame and the candidate frame, and calculating the similarity between corner points; calculating the inter-frame similarity between the current frame and the candidate frame; and taking the pose of the camera underthe candidate frame corresponding to the maximum inter-frame similarity as the pose of the camera under the current frame. According to the method, the influence of illumination and topographic changes on repositioning can be reduced, the precision and efficiency of repositioning are improved, the robustness of a positioning system is enhanced, and the calculated amount is reduced.

Description

technical field [0001] The invention relates to the technical field of relocation, in particular to a camera relocation method based on semantic landmarks. Background technique [0002] Relocation is one of the most common problems in SLAM (simultaneous localization and mapping, real-time positioning and map construction), which solves the problem that robots or drones lose their positioning during the movement for some reason and cannot continue to work. The problem. The relocation of SLAM is defined as: how to find out its own positioning based on the current sensor observation data of the machine when the environmental map is known. There are mainly several situations that lead to the loss of positioning: 1) The machine is moved after it stops working; 2) Due to environmental factors such as terrain and moving objects, the machine observes an environment with a large difference. These situations cause the machine to think it is in a new environment, when in fact it is i...

AI Technical Summary

Problems solved by technology

However, these related semantic SLAM systems are affected by the calculation speed and the positioning accuracy of SLAM itself. These methods using related semantic SLAM systems are basically carried out in ideal scenarios, with high object discrimination and high relocation accuracy and efficiency. Relatively low, the robustness of the positioning system is not high enough

Images