Single-frame image three-dimensional model surface reconstruction method based on deep learning

Abstract

The invention provides a single-frame image three-dimensional model surface reconstruction method based on deep learning, and the method comprises the steps of carrying out the sampling and rendering on a CAD model, and generating the single-frame images under different distances between a model real shape point cloud and different viewpoints; performing feature extraction on the image through a convolutional neural network to obtain the high-level semantics of the two-dimensional image; converting the obtained high-level semantics into the control point coordinates and weight parameters required by NURBS curved surface change in a three-dimensional reconstruction stage through a full-connection neural network module; updating the initialized NURBS model by using the obtained control point coordinates and weight parameters, and gradually performing three-dimensional reconstruction; and training the deep learning model, inputting the training sample into the deep learning model for automatic training to obtain an optimal model parameter, and completing the three-dimensional reconstruction. According to the invention, the three-dimensional reconstruction can be simply and efficiently carried out on the single-frame images, and the reconstructed three-dimensional model has the characteristics of rich details, smooth surface and good overall performance.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a method for surface reconstruction of a three-dimensional model of a single frame image based on deep learning. Background technique [0002] 3D reconstruction refers to the process of establishing a mathematical model of a 3D object in a real scene in a computer, and it is a popular research direction in the field of computer vision. Compared with two-dimensional images, three-dimensional models can provide the overall information of the object and more comprehensively display the characteristics of the object, so it has a wide range of applications in computer animation, human-computer interaction, modern medicine and other fields. [0003] With the rapid development of deep learning, new breakthroughs have been made in the field of 3D reconstruction of single frame images. Researchers use convolutional neural networks to extract features from single-frame images, and reconstru...

AI Technical Summary

Problems solved by technology

However, the traditional setting of control point weights and coordinates requires a large number of initialization calculations and manual fine-tuning, which makes the reconstruction of complex surface 3D models more difficult.

Images