Visual SLAM closed-loop detection method based on target detection

Abstract

The embodiment of the invention relates to a visual SLAM closed-loop detection method based on target detection, and the method comprises the steps of obtaining an RGB image and a depth image for an environment, and carrying out the target detection on the RGB image, and obtaining a target detection result; performing clustering fusion on a key frame determined according to the RGB image and the target detection result by using a DBSCAN density clustering algorithm to obtain a clustering result; performing the spatial position restoration by taking the clustering result as a semantic node in combination with the depth image, and constructing a local semantic topological graph; performing node matching on the local semantic topological graph by using the SURF feature matching and the category of the semantic nodes, and calculating the similarity of the semantic nodes of the local semantic topological graph according to a node matching result; and performing closed-loop judgment based onwhether the similarity of the continuous key frames meets a preset condition or not. According to the method provided by the embodiment of the invention, the errors and omissions possibly existing intarget detection can be eliminated, and the condition of performing environment description by utilizing the target detection result can be improved.

Description

technical field [0001] The present disclosure relates to the technical field of intelligent robot scene recognition, and in particular to a visual SLAM closed-loop detection method based on target detection. Background technique [0002] Simultaneous localization and mapping (SLAM for short) is one of the core issues in mobile robot research technology. Compared with laser sensors, vision sensors have the advantages of delicate perception, low price, smaller volume, and lighter weight. In visual SLAM, the sensor error and the error of the robot's own motion estimation will continue to accumulate. After a long period of motion, a large cumulative error will often occur, which will easily lead to the failure of mapping. Therefore, it is necessary to use the closed-loop detection link to eliminate the influence of accumulated errors. [0003] The loop closure detection problem is essentially a scene recognition problem, and its basic content is to compare the current scene wi...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the above-mentioned problems in the prior art, the present disclosure provides a visual SLAM closed-loop detection method based on target detection, which solves the actual scene of closed-loop detection in the prior art There is a problem that the accuracy is not high enough in the recognition process

Images