Three-dimensional scene reconstruction method and device based on vision SLAM

Abstract

The present invention discloses a three-dimensional scene reconstruction method and device based on vision SLAM. The method comprises the following steps of: vision information obtaining: allowing a mobile robot to freely move in a three-dimensional scene to collect images, and employing an SLAM algorithm to estimate pose information of the mobile robot; point cloud reconstruction: employing an SIFT algorithm to extract feature points in the collected images for matching, employing the SFM algorithm to perform sparse point cloud reconstruction of the matched feature points and the corresponding mobile robot pose information, and performing dense point cloud reconstruction of the reconstructed sparse point cloud; and surface reconstruction: performing surface reconstruction of the reconstructed dense point cloud to complete reconstruction of the three-dimensional scene. The three-dimensional scene reconstruction method and device are high in operation speed, low in requirement for hardware and good in scale invariance. Besides, the three-dimensional scene reconstruction method and device further reconstruct object surface information so as to improve the precision of the reconstructed three-dimensional scene.

Description

technical field [0001] The present invention relates to the technical field of 3D scene reconstruction, in particular to a visual SLAM-based 3D scene reconstruction method and device. Background technique [0002] Computer vision has become a major topic in the field of robotics, and real-time localization and mapping (SLAM) methods have attracted a large number of researchers in the past decade. In terms of research on the monocular vision SLAM system based on the camera sensor device Kinect, Nikolas et al. used Kinect to implement a real-time SLAM system and integrated the system into a read-only memory (ROS). Henry et al. used Kinect to implement an interactive real-time 3D reconstruction system, that is, depth image mapping (RGB-D Mapping). The system selects different key frames for closest point iterative algorithm (ICP) registration according to user needs. , the reconstructed point cloud can well match the top view of the building, but the actual running speed of th...

AI Technical Summary

Problems solved by technology

Due to the high computational complexity of the ICP algorithm, the ICP algorithm accelerated by GPU has high requirements for hardware, which restricts its application range

[0003] However, the 3D scene reconstruction method based on visual SLAM in the prior art only uses the SLAM system to reconstruct the point cloud through feature point matching, which requires high hardware requirements and low operating speed. The surface information is not reconstructed, so the accuracy of the reconstructed 3D scene is poor

Images