Closed-loop detection system and method based on multi-feature fusion

Abstract

The invention discloses a closed-loop detection system and method based on multi-feature fusion. The closed-loop detection system based on multi-feature fusion comprises a data collection module usedfor collecting an image of a scene by using a laser sensor and preprocessing the image of the scene, a geometric feature extraction module used for extracting the geometric structure features of the scene, a visual feature extraction module used for extracting the visual features of the scene, a feature fusion module used for fusing the geometric structure features and the visual features througha data dimension reduction method, and a closed-loop detection module used for carrying out closed-loop detection by using the fused features. According to the closed-loop detection system and methodbased on multi-feature fusion, robust scene feature representation is obtained by fusing the visual features and the geometric structure features of the appearance of the scene, 'perceptual confusion'in scene matching is reduced by making use of the geometric space constraints of feature matching, and the error detection rate is reduced. Therefore, the scheme of the invention is of high expansibility and robustness.

Description

technical field [0001] The invention relates to closed-loop detection technology, in particular to a closed-loop detection system and method based on multi-feature fusion. Background technique [0002] With the improvement of various performances, service robots can complete more and more tasks in people's daily life, such as cleaning, moving objects and so on. In order to complete the task more smoothly, the robot must have a more detailed and accurate perception and understanding of the surrounding environment. [0003] Map representation is the basis of robot positioning and mapping, that is, using some special points, lines, surfaces in the scene or some visual features in the scene image to represent the pose of the robot. By matching and comparing such features, you can Estimate the current pose of the robot. [0004] Loop-closed detection plays an extremely important role in improving the stability of the robot map representation algorithm. The basic definition of c...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art, especially to solve the problem of "perceptual confusion" that a large number of repeated visual scenes such as doors and windows in the indoor environment will cause visual feature matching, and the method based on visual features cannot make full use of the indoor environment The problem of structured and semi-structured features of geometric scenes in

Images