Robust localization and mapping method and system based on fusion of laser and visual information

Abstract

The invention discloses a robust positioning and mapping method and system for fusion of laser and visual information. The invention includes the fusion calculation of non-visual pose T t m ; Estimate the visual pose from the visual image T t v , if the estimation fails, use the non-visual pose T t m Perform non-visual aided relocalization to calculate visual pose T t v ; to the visual pose T t v and non-visual pose T t m Fusion to obtain the fused robot pose T t f And closed-loop detection is performed. After the closed-loop is detected, the closed-loop optimization algorithm based on graph optimization is used to globally optimize the robot pose, thereby obtaining a two-dimensional grid map and a three-dimensional point cloud map. The present invention can fuse various sensor information to solve the problem of loss of visual SLAM tracking, improve the positioning accuracy of the robot, and build an accurate two-dimensional Raster maps and 3D point cloud maps.

Description

technical field [0001] The invention relates to a robot positioning technology for multi-sensor information fusion, in particular to a robust positioning and mapping method and system for laser and visual information fusion. Background technique [0002] In recent years, with the rapid development of robot technology, Simultaneous Localization and Mapping (SLAM), as the basic key technology in the field of robotics, still has some unresolved problems. For SLAM solutions using different sensors, it can be mainly divided into lidar SLAM and visual SLAM. The main sensor used in lidar SLAM is lidar. Lidar has the advantages of high precision, fast speed, and is not affected by ambient light, and can accurately obtain distance and angle information from obstacles. However, two-dimensional laser radar can only obtain information on one plane, and the amount of information is insufficient to detect obstacles outside the laser scanning plane. At present, the price of 3D lidar is g...

AI Technical Summary

Problems solved by technology

However, two-dimensional laser radar can only obtain information on one plane, and the amount of information is insufficient to detect obstacles outside the laser scanning plane.

At present, the price of 3D lidar is generally very expensive. Due to cost considerations, it is mainly used in the field of driverless cars and is not suitable for use in robots.

Thanks to the rapid development of camera technology and computer performance, visual SLAM has gradually emerged many excellent positioning and mapping methods, but the common problem is that there will be cumulative errors in large environments, and the camera moves too fast, the lighting conditions are extreme, and the scene In the case of severe lack of texture features, visual SLAM has problems such as tracking failure, poor positioning accuracy, and poor mapping effect

Images