基于卷积神经网络的航空零件去噪技术  

Point denoising for aviation parts based on convolutional neural network

作　　者：赵安安[1] 郑炜 郭俊刚[1] Zhao An'an;Zheng Wei;Guo Jungang(AVIC Xi'an Aircraft Industry(Group)Company Ltd.,Shaanxi Xi'an,710089,China)

机构地区：[1]航空工业西安飞机工业(集团)有限责任公司,陕西西安710089

出　　处：《机械设计与制造工程》2023年第1期71-74,共4页Machine Design and Manufacturing Engineering

摘　　要：为去除在用三维激光扫描技术扫描航空零部件时,因扫描环境、设备等因素带来的大量零件点云噪声,提出基于卷积神经网络的航空零件去噪技术。首先应用经典卷积神经网络预测点云法线信息,然后以此进一步对点云进行位置更新,从而实现点云去噪。经实验证明,与目前的去噪方法相比,所提方法在去噪方面更具优越性。In order to remove a large number of part point cloud noise caused by scanning environment,equipment and other factors when scanning aviation parts with 3D laser scanning technology,a denoising technology of aviation parts based on convolutional neural network is proposed.Firstly,the classical convolution neural network is used to predict the normal information of the point cloud,and then the point cloud is further updated,so as to realize the point cloud denoising.Experiments show that the proposed method is superior to the current denoising methods in denoising.

关 键 词：计算机辅助设计 法线估计 卷积神经网络 点云去噪 

分 类 号：V261[航空宇航科学与技术—航空宇航制造工程]