Three-dimensional model reconstruction method based on grid deformation

Abstract

The invention provides a three-dimensional model reconstruction method based on grid deformation, which is used for constructing a training sample set and comprises the following steps of: manufacturing discrete view angle pictures of a plurality of models and corresponding three-dimensional point cloud data; setting a deep learning network model based on the graph convolutional neural network, wherein the deep learning network model based on the graph convolutional neural network comprises a discrete view angle feature fusion module and a grid deformation module, and the output of the discrete view angle feature fusion module is connected with the input of the grid deformation module; setting a loss function, and based on the training sample set, training a deep learning network model based on the graph convolutional neural network; and inputting the discrete view angle picture of the to-be-reconstructed object to the trained network model, performing automatic reconstruction of the three-dimensional grid model, and evaluating the precision. According to the method, the discrete view angle picture and the three-dimensional point cloud data set of the object are learned and trained, and automatic three-dimensional grid model reconstruction can be carried out on objects of different types and sizes stably and accurately.

Description

technical field [0001] The invention belongs to the field of three-dimensional reconstruction, in particular to a three-dimensional model reconstruction method based on grid deformation. Background technique [0002] 3D reconstruction is an extremely challenging problem that has been studied in the field of computer vision for decades. Traditional methods based on multi-view geometry, such as many SFM and SLAM algorithms, involve a series of complex processes, including feature extraction, feature matching, matching point triangulation, etc. Therefore, this method is not robust to challenging scenarios where feature extraction or matching cannot be done efficiently. Furthermore, their 3D reconstruction result is usually a sparse reconstruction that cannot be used directly. To overcome these limitations, several learning-based methods have emerged. Using deep visual image features, depth regression estimation is performed on the scene to obtain dense reconstruction results...

AI Technical Summary

Problems solved by technology

This kind of inferring the 3D model of an object from only one RGB image is undoubtedly very attractive, but the above methods are inherently limited by a single view

Because the geometric prior obtained from an image from only one angle is often too limited to accurately reconstruct the geometry of objects not seen in this image

Images