Autonomous unmanned system position identification and positioning method based on sequence image features

Abstract

The invention discloses an autonomous unmanned system position identification and positioning method based on sequence image features. Firstly, to-be-image features are extracted through an improved convolutional neural network model, the obtained depth features have high illumination invariance and viewing angle invariance, and the robustness of the algorithm for scene condition changes and robotviewing angle changes is enhanced; then, a difference measurement method based on an image sequence is adopted, constraints are effectively provided for position recognition of adjacent frames, and the recognition accuracy is improved. Secondly, an approximate nearest neighbor search method is used, so that the calculated amount of sequence search is greatly reduced, and the use efficiency in a large-scale environment is improved. Finally, through a method of dynamically updating candidate matching, omission caused in sequence search is effectively reduced, and the error-tolerant rate of thealgorithm is improved. A visual position recognition algorithm has the outstanding advantages of being high in robustness, high in efficiency, capable of adapting to various scenes and the like.

Description

technical field [0001] The invention belongs to the field of mobile robots, and relates to a position recognition and positioning method of an autonomous unmanned system based on sequence image features. Background technique [0002] Realizing long-term autonomous navigation and positioning of robots in a dynamically changing environment is one of the main research difficulties and hotspots in mobile robot technology. How to perform efficient position recognition in long-term and large-scale motion environments has become an urgent need to solve The problem. Vision-based position recognition technology retrieves and matches the current image acquired by the robot with the reference image in the map to determine the current position of the robot in the map. When a robot moves for a long time in a large-scale scene, it is in a dynamically changing environment. Affected by factors such as illumination, seasons, weather, occluders, moving objects, and shooting angles, the appea...

AI Technical Summary

Problems solved by technology

However, CNN-based image feature descriptors also have a high dimensionality, which tends to cause a large amount of calculation when performing similarity measurement. Usually, certain dimensionality reduction and optimization are performed on it before subsequent operations.

In addition, the map obtained by moving in a large-scale scene will be relatively large, and it will consume a lot of time when performing retrieval tasks

Images