基于三维激光点云处理的建筑材料自动识别研究OA
Automatic Identification of Building Materials Based on 3D Laser Point Cloud Processing
建筑信息模型的发展对于数字化建模技术提出了更高要求,这也使得针对如何在密集、无序且特征各异的点云样本中准确识别出不同类型的建筑材料这一问题,提出基于三维激光点云处理的建筑材料自动识别方法.通过三维激光扫描获得建筑材料点云数据,利用深度学习下的点云分割网络,对原始点云进行去噪和下采样处理,实现对建筑材料数据的多源融合分析.在此基础上,根据点云数据与目标材料之间的反射强度,定义自动识别模型,并求解其损失函数,完成基于三维激光点云处理的建筑材料自动识别方法设计.针对包含 5 种常见材料的小型点云数据样本集合进行实验分析,结果表明所提出方法在建筑材料识别任务上的平均交并比超过 80%,相较于其他实验方法,三维激光点云处理技术能够提升建筑材料的识别精度;对透射性、反射性较强的玻璃材料而言,应用性能优异的识别方法依然不能获得高准确性的识别结果,表明该类型材料的识别难度较大.
Advances in Building Information Modelling(BIM)have led to increasingly stringent requirements for digital modelling techniques.This has made the accurate identification of various building materials from dense,unordered and feature-diverse point cloud samples an urgent challenge.This study therefore proposes an automatic identification method for building materials based on 3D laser point cloud processing to address this issue.Building material point cloud data are acquired via 3D laser scanning.A deep learning-based point cloud segmentation network is then employed to denoise and downsample the raw point clouds.This enables multi-source fusion analysis of building material data.On this basis,an automatic recognition model is defined according to the reflection intensity between the point cloud data and the target material,and its loss function is subsequently solved to complete the design of the identification method.Experimental analysis was conducted on a small-scale point cloud dataset comprising five common material types.The results show that the proposed method achieves an average intersection over union(Io U)of over 80%for building material recognition tasks,demonstrating that 3D laser point cloud processing technology can effectively improve recognition accuracy compared to alternative methods.However,even high-performance recognition methods still fail to yield highly accurate results for glass materials,which are characterized by strong projection and reflection properties,indicating the inherent difficulty in identifying such materials.
纵瑞芳;王颖
安徽工业经济职业技术学院 地质与建筑工程学院,安徽 合肥 230051安徽新华学院 教务处,安徽 合肥 230088
建筑与水利
三维激光点云处理建筑材料自动识别深度学习反射强度
3D laserpoint cloud processingbuilding materialsautomatic identificationdeep learningreflection intensity
《徐州工程学院学报(自然科学版)》 2026 (3)
30-36,7
安徽省高校自然科学研究项目(2023AH0526752023AH052673)
评论