摘要
在物体的建模过程中,不可避免的存在遮挡、抖动等情况,这造成了三维数据模型信息丢失、模型结构残缺不全的现象。为此,本文讨论了一种处理残缺三维模型的办法,基于深度学习的模型修复。具体体现在两个方面,一方面,本文网络结构用于计算重建误差,从而保证输入输出的一致性,使得输出的模型更加的真实自然;另一方面,网络在提高稳定性的基础上,能够预测得到残缺模型的完整结构,从而使得输出的模型具有更好的视觉效果。和其他实验方法对比,本文的方法能够得到补全程度更高的精致准确输出模型。
In the process of object modeling,there are unavoidable situations such as occlusion and jitter,which cause the loss of three-dimensional data model information and incomplete model structure.Requires predictive filling of missing parts of the model.To this end,this article discusses a method for dealing with incomplete three-dimensional data models.It is embodied in two aspects,on the one hand,the network structure in this paper is used to calculate the reconstruction error,thereby ensuring the consistency of the input and output,and making the output model more real and natural.On the other hand,on the basis of improving the stability,the network can predict the complete structure of the incomplete model,so that the output model has better visual effects.Compared with other experimental methods,the method in this paper can obtain a more accurate and accurate output model with a higher degree of completion.
作者
贝子勒
赵杰煜
BEI Zi-le;ZHAO Jie-yu(Faculty of Electrial Engineering and Computer,NingBo University,Ningbo 315211,China)
出处
《无线通信技术》
2020年第2期6-11,共6页
Wireless Communication Technology
基金
国家自然科学基金(61571247)
浙江省自然科学基金(LZ16F03001)
浙江省国际合作项目(2013C2407)。
关键词
深度学习
三维点云
三维修复
deep learing
three-dimensional point cloud
three-dimensional repair
