Natural landscape multi-view three-dimensional reconstruction method based on deep learning

Abstract

The invention provides a natural landscape multi-view three-dimensional reconstruction method based on deep learning, and the method comprises the steps: obtaining a multi-view image set of a natural landscape, and carrying out the preprocessing of a two-dimensional image in the multi-view image set; constructing a multi-scale feature extraction network, and performing feature extraction on the preprocessed two-dimensional image by using the trained multi-scale feature extraction network to obtain target key features; inputting the target key features into a learning-based patch matching iteration model to carry out iterative calculation of pixel depth matching, and outputting a corresponding depth map after the iterative calculation is finished; the method comprises the steps of obtaining a depth image and a source image, inputting the obtained depth image and the source image into a depth residual network for optimization to obtain an optimized final depth image, and constructing an object three-dimensional model according to the optimized final depth image to obtain a stereoscopic vision image of a natural landscape. Therefore, the obtained depth map is more complete and accurate, and the local detail precision of the landscape model is higher.

Description

technical field [0001] The invention relates to the technical field of three-dimensional reconstruction, in particular to a multi-view natural landscape based on deep learning Figure 3 dimensional reconstruction method. Background technique [0002] Metaverse technology is a virtual reality fusion scene framework that integrates virtual reality, game engine, mobile Internet, blockchain, etc., providing a highly immersive interactive experience. Applying Metaverse technology to real cultural tourism scenarios, improving the innovative development ideas of cultural tourism products, and protecting the intellectual property rights of cultural tourism digital products have very important theoretical significance and practical value for the development of the tourism industry. The key technology of constructing a natural landscape scene integrating virtual and real is to display the image data through multi-view Figure 3 3D reconstruction techniques create simulated virtual sce...

AI Technical Summary

Problems solved by technology

However, due to the deep regularization processing of the 3D convolutional neural network, in large-scale and high-resolution scenes, MVSNet is also subject to the limitation of video memory resources.

The traditional method is difficult to deal with specular reflection, texture, etc., the integrity of the reconstruction is low, and the speed is slow, and the environmental factors of the natural landscape model reconstruction are relatively large, the feature extraction is insufficient, and the parameters are designed in advance and cannot be self-adapted. For specific scene effects, not universally applicable

Images