Laser SLAM (simultaneous localization and mapping) closed loop detection method based on deep learning

Abstract

The invention discloses a laser SLAM (simultaneous localization and mapping) closed loop detection method based on deep learning. According to the method, an SLAM closed loop detection problem is converted into an SLAM data sample search problem; a deep Hash network for carrying out Hash coding on laser point cloud samples is established innovatively; and sample similarity calculation is carried out based on the Hash coding, thereby realizing similar sample rapid search and SLAM closed loop detection. According to the established deep Hash network, through utilization of the advantages of the deep learning, binary coding more reliable than a traditional coding technology can be obtained, so the search accuracy is greatly improved. A new point cloud characteristic extraction algorithm is designed, through application of the algorithm, on the basis of carrying out space division and projection on point clouds, a characteristic graph is established, laser point cloud sample data is reliably described, and effective docking with the deep Hash network is realized. According to the established deep Hash network, through offline training, a trained model can satisfy real-time coding demands.

Description

technical field [0001] The invention relates to the field of mobile mapping and autonomous navigation, in particular to a laser SLAM closed-loop detection method based on deep learning. technical background [0002] SLAM (simultaneous localization and mapping), real-time positioning and map construction. SLAM was first proposed in 1988. Due to its important theoretical and application value, it is considered by many scholars to be the key to realizing a truly fully autonomous mobile robot. The SLAM problem can be described as: the robot starts to move from an unknown position in an unknown environment, locates itself according to position estimation and maps during the movement process, and builds an incremental map on the basis of its own positioning to realize the autonomous positioning and positioning of the robot. navigation. At present, the sensors used in SLAM are mainly divided into two categories, one is lidar, and the other is camera. The corresponding SLAM techno...

AI Technical Summary

Problems solved by technology

However, the computational efficiency of the ICP algorithm is low, and the feature extraction, coding, and classification of the bag-of-words technology are usually based on the artificially constructed feature space, and its adaptability is limited. Poor reliability, unable to meet high-precision laser SLAM positioning and composition requirements

Images