基于卷积神经网络的非织造材料三维点云重建  

3D point cloud reconstruction of nonwoven materials based on convolution neural networks

作　　者：陈旭旭 王荣武[1] Chen Xuxu;Wang Rongwu(College of Textiles,Donghua University,Shanghai 201620,China)

机构地区：[1]东华大学纺织学院,上海201620

出　　处：《产业用纺织品》2021年第4期50-57,共8页Technical Textiles

基　　金：国家自然科学基金(61771123)。

摘　　要：针对传统的从聚焦点获取深度重建非织造材料三维点云模型带来的背景噪声问题,提出了一种用于检测图像中聚焦点的卷积神经网络IPDNet。试验对比分析表明,该网络能够较好地剔除图像背景的噪声点,在重建非织造材料三维点云模型中具有良好的效果。In terms of the noise problems caused by the traditional nonwoven 3D point cloud model obtained from the focus point depth reconstruction,a convolutional neural network IPDNet for detecting the focus point in the image was proposed.Experimental comparative analysis proved that the network would well eliminate the noise points in the background of the focused image,and had a good effect in reconstructing the nonwoven 3D point cloud model.

关 键 词：非织造材料 三维点云 光学显微镜 卷积神经网络 多焦面图像 

分 类 号：TS171[轻工技术与工程—纺织材料与纺织品设计]