Visual SLAM loopback verification system based on consistency cluster

Abstract

The invention relates to a visual SLAM loopback verification system based on a consistency cluster. The system comprises a front-end visual odometer module which is used for extracting features of input image frames, determining the position relation between adjacent image frames, and constructing a local map; a loopback detection module is used for judging that the image frames are at the same position on the basis of the image frames calculated by the front-end visual odometer and the similarity, namely, judging loopback; a rear-end nonlinear optimization module is used for globally optimizing the local map input by the front-end visual odometer module and the loopback constraint output by the loopback detection module through a graph optimization method to obtain a globally consistent track and map. Compared with the prior art, the method has the advantages of high reliability, good accuracy, low resource consumption and the like.

Description

technical field [0001] The invention relates to the fields of artificial intelligence and robot control, in particular to a visual SLAM loopback verification system based on consistency clusters. Background technique [0002] In recent years, human society has developed rapidly in the direction of intelligence and industrialization. Under this environment, a large number of research and development projects and robotic products have emerged at home and abroad, involving various fields of life. Mobile robots are one of them. Robots of concern. For example, Google's self-driving car project, which aims to develop driverless cars and help drivers complete operations such as starting and driving the vehicle. The project has entered the stage of actual road testing and completed a driving distance of over one million kilometers. In addition, other mobile robots, including household cleaning robots, restaurant delivery robots, security robots, and large-scale freight automatic g...

AI Technical Summary

Problems solved by technology

[0005] After the visual SLAM loop detection algorithm judges that the robot has reached the same position through the image frame, this information is added as an optimization constraint to the back-end nonlinear optimization process to help the system obtain a more accurate map and more accurate positioning. However, it is only based on The reliability of the single loop closure hypothesis determined by the similarity of the two image frames is not high enough, which may cause some wrong constraint conditions to be added, resulting in wrong map estimation. Therefore, further research is needed on the positioning accuracy of the visual SLAM loop closure verification system

Images