Robot indoor mapping method and system based on vision and laser slam

Abstract

The invention discloses a robot indoor mapping method based on vision and laser slam. The method comprises the following steps that: a vision sensor acquires a left view and a right view to generate afeature point cloud; pose transformation estimation of a robot based on vision is calculated; laser data are obtained, robot pose transformation estimation based on laser is calculated, and a confidence coefficient of map grids is updated; a tile type mapping method is adopted and continuous laser point cloud data are utilized to generate sub-graphs; obtained feature points matches the sub-graphs; and the robot pose in each sub-graph is updated through graph optimization to form a final mapping result. According to the method, a visual and laser combined mapping method is adopted; on the onehand, the defect that a visual algorithm is greatly influenced by illumination is overcome; on the other hand, closed loops are detected through descriptor attributes of feature points, so that a problem that the success rate of laser closed loop detection is too low is solved and the closed loop detection efficiency is improved.

Description

technical field [0001] The invention relates to the field of inspection robots, in particular to a method and system for indoor mapping of a robot based on vision and laser slam. Background technique [0002] The quality of the indoor navigation environment directly affects the completion of various inspection tasks of the robot. To complete the creation of the map, it is necessary to obtain the information of the surrounding environment first, which requires the use of sensors. At present, the research hotspots of robot mapping are generally concentrated on two types of sensors, one is the laser sensor, the laser sensor has high resolution, high ranging accuracy, strong anti-interference ability, good detection performance, and is not affected by light. The reason is that the obtained information does not have semantic information, the registration accuracy is not high, and the closed-loop detection often fails during the mapping process. Another type of visual sensor can...

AI Technical Summary

Problems solved by technology

At present, the research hotspots of robot mapping are generally concentrated on two types of sensors, one is the laser sensor, the laser sensor has high resolution, high ranging accuracy, strong anti-interference ability, good detection performance, and is not affected by light. The obtained information does not have semantic information, the registration accuracy is not high, and the closed-loop detection often fails during the mapping process

Another type of visual sensor, which can obtain rich texture information, has the advantage of accurate matching, but has the disadvantage of being greatly affected by illumination changes

Images