Three-dimensional reconstruction method based on attention mechanism and monocular multi-view angle

Abstract

The invention discloses a three-dimensional reconstruction method based on an attention mechanism and monocular multi-view angles. The method comprises the steps that S1, a to-be-detected scene is shot through a camera, and image data of the to-be-detected scene is collected; s2, performing order marking on the image data, and performing feature point extraction and feature point matching in sequence to obtain feature point matching pairs; s3, calibrating the camera to obtain camera pose information, camera internal reference and to-be-measured scene structure information, and performing sparse point cloud reconstruction; s4, inputting the camera pose information, the camera internal reference, the to-be-measured scene structure information and the image into a monocular multi-view reconstruction network of a preset attention mechanism to obtain a plurality of depth estimation maps of the to-be-measured scene; and S5, performing depth fusion on the plurality of depth estimation maps to obtain a dense point cloud model. The method is suitable for reconstruction of any scene, and is more accurate in reconstruction result, high in point cloud integrity, simple in process, reliable in reconstruction and high in adaptability.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a three-dimensional reconstruction method based on an attention mechanism and monocular multi-view angles. Background technique [0002] In recent years, deep learning technology has shined in various tasks of two-dimensional image processing, and the accuracy in various data sets far exceeds the results of traditional methods. With the increasing demand for 3D vision and 3D structural data, researchers began to apply deep learning technology to 3D vision technology. The 3D reconstruction problem is a classic computer vision problem, so combining it with neural networks to obtain higher-precision reconstruction results has become a widely researched problem. [0003] Compared with lidar and other methods, the equipment required for image-based 3D reconstruction technology is simple and cheap, and the obtained models are more widely used. The technology of 3D reconstr...

AI Technical Summary

Problems solved by technology

[0004] The technical purpose of the present invention is to provide a 3D reconstruction method based on the attention mechanism and monocular multi-view angles to solve the technical problems of insufficient accuracy and completeness in 3D reconstruction

Images