A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Rock Discontinuities Identification from 3D Point Clouds Using Artificial Neural Network
https://orcid.org/0000-0003-4390-6841
Abstract Rock discontinuities fundamentally impact the mechanical and hydraulic behaviors of a rock mass, and thus it is a critically important task to characterize the geometrical parameters of these rock discontinuities. To measure the discontinuity orientation more accurately and efficiently, two well-known point clouds were taken as cases (a cube and a road cut), and an artificial neural network (ANN)-an machine learning algorithm-was established to identify discontinuities from point clouds through learning a small number of training samples, which had been manually selected from the raw point clouds. Four attributes associated with geometrical features of point clouds were specified as input parameters, namely, point XYZ-coordinates, point normal, point curvature, and point density. Two main groups-discontinuity and non-discontinuity-were produced in the output layer, and the number of the discontinuity groups greatly depended on the sets of discontinuities in the real situation. Using principal component analysis (PCA) and density-based spatial clustering of applications with noise (DBSCAN), single discontinuities were extracted from the group discontinuities which were obtained using ANN, and the corresponding orientations were calculated. The results obtained with the proposed method in this study matched the field surveys and results calculated by a modified region-growing algorithm. The computational efficiency was significantly enhanced using the proposed method, only taking several seconds to process a huge data. More importantly, the accuracy of discontinuity detection was greatly improved by specifying the noise data as the non-discontinuity groups during training samples selection in ANN. The ANN approach does not require the engineers have a strong professional background in computer programming, which simplified the detection and characterization process of rock discontinuity. Furthermore, an APP-named DisDetANN-was developed to implement the rock discontinuity detection based on the proposed ANN model, and the full code of the DisDetANN has been freely shared on GitHub.
Highlights An artificial neural network was created by machine learning to detect group discontinuities from point clouds.Point coordinate, normal, curvature, and density were considered in input layers of the artificial neural network.A clustering algorithm was employed to subdivide group discontinuities into single discontinuities.Both efficiency and accuracy of the discontinuity detection was improved by the proposed approach.An APP and full codes of the proposed method were freely made available to the engineering community.
Rock Discontinuities Identification from 3D Point Clouds Using Artificial Neural Network
https://orcid.org/0000-0003-4390-6841
Abstract Rock discontinuities fundamentally impact the mechanical and hydraulic behaviors of a rock mass, and thus it is a critically important task to characterize the geometrical parameters of these rock discontinuities. To measure the discontinuity orientation more accurately and efficiently, two well-known point clouds were taken as cases (a cube and a road cut), and an artificial neural network (ANN)-an machine learning algorithm-was established to identify discontinuities from point clouds through learning a small number of training samples, which had been manually selected from the raw point clouds. Four attributes associated with geometrical features of point clouds were specified as input parameters, namely, point XYZ-coordinates, point normal, point curvature, and point density. Two main groups-discontinuity and non-discontinuity-were produced in the output layer, and the number of the discontinuity groups greatly depended on the sets of discontinuities in the real situation. Using principal component analysis (PCA) and density-based spatial clustering of applications with noise (DBSCAN), single discontinuities were extracted from the group discontinuities which were obtained using ANN, and the corresponding orientations were calculated. The results obtained with the proposed method in this study matched the field surveys and results calculated by a modified region-growing algorithm. The computational efficiency was significantly enhanced using the proposed method, only taking several seconds to process a huge data. More importantly, the accuracy of discontinuity detection was greatly improved by specifying the noise data as the non-discontinuity groups during training samples selection in ANN. The ANN approach does not require the engineers have a strong professional background in computer programming, which simplified the detection and characterization process of rock discontinuity. Furthermore, an APP-named DisDetANN-was developed to implement the rock discontinuity detection based on the proposed ANN model, and the full code of the DisDetANN has been freely shared on GitHub.
Highlights An artificial neural network was created by machine learning to detect group discontinuities from point clouds.Point coordinate, normal, curvature, and density were considered in input layers of the artificial neural network.A clustering algorithm was employed to subdivide group discontinuities into single discontinuities.Both efficiency and accuracy of the discontinuity detection was improved by the proposed approach.An APP and full codes of the proposed method were freely made available to the engineering community.
Rock Discontinuities Identification from 3D Point Clouds Using Artificial Neural Network
https://orcid.org/0000-0003-4390-6841
Abstract Rock discontinuities fundamentally impact the mechanical and hydraulic behaviors of a rock mass, and thus it is a critically important task to characterize the geometrical parameters of these rock discontinuities. To measure the discontinuity orientation more accurately and efficiently, two well-known point clouds were taken as cases (a cube and a road cut), and an artificial neural network (ANN)-an machine learning algorithm-was established to identify discontinuities from point clouds through learning a small number of training samples, which had been manually selected from the raw point clouds. Four attributes associated with geometrical features of point clouds were specified as input parameters, namely, point XYZ-coordinates, point normal, point curvature, and point density. Two main groups-discontinuity and non-discontinuity-were produced in the output layer, and the number of the discontinuity groups greatly depended on the sets of discontinuities in the real situation. Using principal component analysis (PCA) and density-based spatial clustering of applications with noise (DBSCAN), single discontinuities were extracted from the group discontinuities which were obtained using ANN, and the corresponding orientations were calculated. The results obtained with the proposed method in this study matched the field surveys and results calculated by a modified region-growing algorithm. The computational efficiency was significantly enhanced using the proposed method, only taking several seconds to process a huge data. More importantly, the accuracy of discontinuity detection was greatly improved by specifying the noise data as the non-discontinuity groups during training samples selection in ANN. The ANN approach does not require the engineers have a strong professional background in computer programming, which simplified the detection and characterization process of rock discontinuity. Furthermore, an APP-named DisDetANN-was developed to implement the rock discontinuity detection based on the proposed ANN model, and the full code of the DisDetANN has been freely shared on GitHub.
Highlights An artificial neural network was created by machine learning to detect group discontinuities from point clouds.Point coordinate, normal, curvature, and density were considered in input layers of the artificial neural network.A clustering algorithm was employed to subdivide group discontinuities into single discontinuities.Both efficiency and accuracy of the discontinuity detection was improved by the proposed approach.An APP and full codes of the proposed method were freely made available to the engineering community.
Rock Discontinuities Identification from 3D Point Clouds Using Artificial Neural Network
Ge, Yunfeng (author) / Cao, Bei (author) / Tang, Huiming (author)
2022
Article (Journal)
Electronic Resource
English
BKL:
38.58
Geomechanik
/
56.20
Ingenieurgeologie, Bodenmechanik
/
38.58$jGeomechanik
/
56.20$jIngenieurgeologie$jBodenmechanik
RVK:
ELIB41
| Elsevier | 2024