Vision-based topological map generation method

Abstract

The invention discloses a vision-based topological map generation method, comprising: acquiring image information, extracting image frames at a predetermined frequency based on the image information; obtaining key frames based on image frame preprocessing, judging whether a loop is currently generated, and correcting the currently existing loop if a loop is generated If there is a topological map, if the key frame is the first frame of image in this mapping operation, set a topological node at its position; there are two ways to distinguish other topological points: one is to compare the image information with the existing topological map The image of the newly added topological node in the image, judge whether the descriptor similarity of the two images is lower than the first threshold, if it is lower than the first threshold, add the location of the image of the key frame as a new topological node to the topological map, store current frame. The second is to judge whether the angle information of several consecutive frames is greater than the second threshold relative to the previous topological node, and if it is larger, the position of the image of the key frame is added to the topological map as a new topological node, and the current frame is stored.

Description

technical field [0001] The invention belongs to the technical field of computer vision, in particular to a method for generating topological maps based on vision. Background technique [0002] Unmanned systems, such as self-driving unmanned vehicles and unmanned aerial vehicles, use the built-in sensors of the system to acquire and describe the driving state of the system, and use the obtained information to complete autonomous navigation tasks. The sensors used in general unmanned systems include lasers, radars, cameras, and IMUs. Due to the high price of laser and radar, the price of camera is relatively low, and the navigation technology based on computer vision technology has been widely researched and applied. [0003] Simultaneous Localization and Mapping (SALM) technology refers to a system equipped with specific sensors that estimates its own state through motion and builds a map including surrounding environment information during the motion. When we primarily use...

AI Technical Summary

Problems solved by technology

However, in the existing methods, the generation of topological maps lacks effective logic, and the efficiency of the generated nodes for space representation is relatively low.

Images