摘要
岩体点云滤波是岩体三维重建的关键环节。针对岩体点云环境,提出一种基于多维度特征和多层神经网络的植被滤波方法。该方法首先计算点云中每一点的多维度特征作为特征输入;然后利用多层神经网络构建分类器实现对岩体点云数据的植被滤波过程。分析多维度特征的可用性,并通过不同的实验过程筛选最优网络模型参数。与其他分类器相比,本算法精度较高,能够更好地应用于岩体点云植被滤波领域。
Filtering on rock mass point clouds is an important step in 3D rock mass reconstruction.This work focuses on rock mass point clouds and we propose a vegetation filtering method based on multi-dimensionality features and MLP(multi-layer perceptron).This method firstly calculates multi-dimensionality features for each point.Then,MLP is used for training the classifier,which can be applied in vegetation filtering.We analyze the availability of multi-dimensionality features and select the best MLP model through different experimental processes.The experimental results show that the proposed method has a higher precision than other classifiers and it can be better applied in the field of rock mass point cloud vegetation filtering.
作者
胡亮
肖俊
王颖
HU Liang;XIAO Jun;WANG Ying(School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China)
出处
《中国科学院大学学报(中英文)》
CSCD
北大核心
2020年第3期345-351,共7页
Journal of University of Chinese Academy of Sciences
基金
国家自然科学基金(61471338)
中国科学院青年促进会(2015361)
中国科学院前沿科学重点研究项目(QYZDY-SSW-SYS004)
北京市科技新星计划(Z171100001117048)
北京市科技计划课题(Z181100003818019)资助。
关键词
岩体点云
植被滤波
维度特征
多层神经网络(MLP)
rock mass point clouds
vegetation filtering
dimensionality features
multi-layer perceptron(MLP)