With the rapid development of 3D digital photography and 3D digital scanning devices, massive amount of point samples can be generated in acquisition of complex, real-world objects, and thus create an urgent need for ...With the rapid development of 3D digital photography and 3D digital scanning devices, massive amount of point samples can be generated in acquisition of complex, real-world objects, and thus create an urgent need for advanced point-based processing and editing. In this paper, we present an interactive method for blending point-based geometries by dragging-and- dropping one point-based model onto another model’s surface metaphor. We first calculate a blending region based on the polygon of interest when the user drags-and-drops the model. Radial basis function is used to construct an implicit surface which smoothly interpolates with the transition regions. Continuing the drag-and-drop operation will make the system recalculate the blending regions and reconstruct the transition regions. The drag-and-drop operation can be compound in a constructive solid geometry (CSG) manner to interactively construct a complex point-based model from multiple simple ones. Experimental results showed that our method generates good quality transition regions between two raw point clouds and can effectively reduce the rate of overlapping during the blending.展开更多
为解决现有算法在处理不同特征点云时存在普适性不足、效率低下、难以应用于工程实际的问题,提出一种基于神经网络的岩体结构面智能识别方法,具体包括4个步骤。首先,对原始点云进行标准化预处理操作,并人工选取具有代表性的特征区域,以...为解决现有算法在处理不同特征点云时存在普适性不足、效率低下、难以应用于工程实际的问题,提出一种基于神经网络的岩体结构面智能识别方法,具体包括4个步骤。首先,对原始点云进行标准化预处理操作,并人工选取具有代表性的特征区域,以构建高质量的训练样本集;其次,采用CFSFDP(clustering by fast search and find of density peaks)聚类算法为样本生成标签;再次,构建并训练多层感知机(multilayer perceptron,MLP)模型和多层卷积神经网络(multi-layer convolutional neural network,MCNN)模型,输入全尺度点云的点法向量进行结构面粗识别,并对2种模型进行比选分析;最后,使用HDBSCAN(hierarchical density-based spatial clustering of applications with noise)算法对分类结果进行细化与产状计算。结果表明:1)采用多层感知机模型处理简单结构面时具有较高的处理速度,而卷积神经网络模型在处理复杂、非均匀点云时展现出更高的分类精度。2)与聚类方法相比,该方法计算时间提升25%~50%,能够有效解决传统算法无法适用于不同复杂点云的问题,且具有很强的鲁棒性。展开更多
基金Project supported by the National Natural Science Foundation of China (Nos. 60473106 and 60333010)the Program for Chang-jiang Scholars and Innovative Research Team in University (No. IRT0652), China
文摘With the rapid development of 3D digital photography and 3D digital scanning devices, massive amount of point samples can be generated in acquisition of complex, real-world objects, and thus create an urgent need for advanced point-based processing and editing. In this paper, we present an interactive method for blending point-based geometries by dragging-and- dropping one point-based model onto another model’s surface metaphor. We first calculate a blending region based on the polygon of interest when the user drags-and-drops the model. Radial basis function is used to construct an implicit surface which smoothly interpolates with the transition regions. Continuing the drag-and-drop operation will make the system recalculate the blending regions and reconstruct the transition regions. The drag-and-drop operation can be compound in a constructive solid geometry (CSG) manner to interactively construct a complex point-based model from multiple simple ones. Experimental results showed that our method generates good quality transition regions between two raw point clouds and can effectively reduce the rate of overlapping during the blending.
文摘为解决现有算法在处理不同特征点云时存在普适性不足、效率低下、难以应用于工程实际的问题,提出一种基于神经网络的岩体结构面智能识别方法,具体包括4个步骤。首先,对原始点云进行标准化预处理操作,并人工选取具有代表性的特征区域,以构建高质量的训练样本集;其次,采用CFSFDP(clustering by fast search and find of density peaks)聚类算法为样本生成标签;再次,构建并训练多层感知机(multilayer perceptron,MLP)模型和多层卷积神经网络(multi-layer convolutional neural network,MCNN)模型,输入全尺度点云的点法向量进行结构面粗识别,并对2种模型进行比选分析;最后,使用HDBSCAN(hierarchical density-based spatial clustering of applications with noise)算法对分类结果进行细化与产状计算。结果表明:1)采用多层感知机模型处理简单结构面时具有较高的处理速度,而卷积神经网络模型在处理复杂、非均匀点云时展现出更高的分类精度。2)与聚类方法相比,该方法计算时间提升25%~50%,能够有效解决传统算法无法适用于不同复杂点云的问题,且具有很强的鲁棒性。
