摘要
阐述基于深度学习的多视图立体匹配三维重建算法,包括特征提取、微分单应性变换计算代价体、代价体正则化、深度图与重建点云.三维重建有利于学者们对场景做出观察、感受、学习和研究.
Abstract
This paper expounds a multi view stereo matching 3D reconstruction algorithm based on deep learning,including feature extraction,differential homography transformation to calculate cost volume,cost volume regularization,depth map and reconstructed point cloud.3D reconstruction is beneficial for scholars to observe,feel,learn,and study the scene.
维普
万方数据