Source-linked AI summary
A region-growing approach for automatic outcrop fracture extraction from a three-dimensional point cloud
Xin Wang, Lejun Zou, Xiaohua Shen, Yupeng Ren, Yi Qin
TL;DR
Manual fracture surveys are labor-intensive and may not represent a study region, while automatically extracting the full extent of each fracture from 3D outcrop point clouds remains a challenge. The paper applies a region-growing method using local surface features to extract fractures and estimate their orientations. Tests report automatic, accurate full-extent extraction and fracture data quality as good as or better than a manual survey.
Problem
Manual fracture surveys are time-consuming and can produce datasets too small to represent the study region, while existing point-cloud methods have not reliably extracted each fracture’s full extent automatically.
Source: §1 · p. 2Method
A region-growing algorithm uses local surface normals and curvature, with seed selection and growth criteria that account for overall fracture occurrence, to extract fractures and estimate their orientations.
Source: §1 · p. 4, §3 · p. 6, §5 · p. 21Results
The tests report automatic, accurate extraction of each individual fracture’s full extent, with fracture-data quality as good as or better than manual field-survey data.
Source: §5 · p. 21Takeaways & Limitations
The method can also extract many small fracture faces for which conventional manual measurements cannot be obtained.
Source: §4.2 · p. 14Possible outlier clusters may occur in the extracted fracture sets, and their nature and removal require further study.
Source: §4.2 · p. 17
Abstract
from arXiv · showhide
Conventional manual surveys of rock mass fractures usually require large amounts of time and labor; yet, they provide a relatively small set of data that cannot be considered representative of the study region. Terrestrial laser scanners are increasingly used for fracture surveys because they can efficiently acquire large area, high-resolution, three-dimensional (3D) point clouds from outcrops. However, extracting fractures and other planar surfaces from 3D outcrop point clouds is still a challenging task. No method has been reported that can be used to automatically extract the full extent of every individual fracture from a 3D outcrop point cloud. In this study, we propose a method using a region-growing approach to address this problem; the method also estimates the orientation of each fracture. In this method, criteria based on the local surface normal and curvature of the point cloud are used to initiate and control the growth of the fracture region. In tests using outcrop point cloud data, the proposed method identified and extracted the full extent of individual fractures with high accuracy. Compared with manually acquired field survey data, our method obtained better-quality fracture data, thereby demonstrating the high potential utility of the proposed method.
1. Introduction
Manual fracture surveys are labor-intensive and may not represent the study region, while extracting complete individual fractures from 3D outcrop point clouds remains challenging. The proposed region-growing algorithm aims to extract each fracture’s full extent automatically and estimate its orientation.
1. Introduction
- Manual surveys are time-consuming and usually yield a relatively small dataset, while TLS can acquire high-resolution 3D outcrop point clouds.Fracture data can support structural and geomechanical analysis, flow modeling, reservoir characterization, and rock mass classification.Source: §1 · p. 2
- TLS point clouds can be transformed into geographic coordinate systems for topographic feature extraction and planar-surface orientation estimation.A GPS receiver module in a typical TLS enables this transformation.Source: §1 · p. 3
- Prior extraction methods either failed to capture full individual fracture surfaces or required human supervision.An automatic method also split many individual fractures into several planar structures rather than extracting their full extent.Source: §1 · pp. 3–4
- The proposed region-growing algorithm targets automatic extraction of each individual outcrop fracture’s full extent and estimates its orientation.Its seed-point selection and growth criteria consider the overall fracture occurrence.Source: §1 · p. 4
2. Study area and database
The study examines a small sandstone road-cut outcrop scanned with a terrestrial laser scanner. A central, less obstructed part of the resulting point cloud was used for testing, alongside manual fracture-orientation measurements.
2. Study area and database
- The study site is a roughly 30 m2 road-cut outcrop of thin- to thick-layered sandstone in Nanbaoxiang, Chengdu, China.A RIEGL VZ-1000 terrestrial laser scanning system was used to acquire the point cloud.Source: §2 · p. 4
- The test used a central, less vegetated and less obstructed part of a point cloud containing about 21 million points, with average spacing below 1 cm.The central test area was selected because it had less vegetation and fallen stone; 65 manual dip-direction and dip-angle measurements were also collected for comparison.Source: §2 · p. 5
- Figure 1 shows the layered road-cut slope, with a white rectangle marking the test area and a red notebook providing scale.Source: §2 · p. 5, Figure 1 · p. 5
3. Methodology
The method segments fractures directly in point clouds using local normals and curvature, with region-growing criteria chosen to incorporate the fracture’s overall geometry. After growth, it estimates fracture orientation from a fitted plane.
3. Methodology
- The algorithm works directly on the point cloud, using local surface normals and curvature to grow fracture regions and then estimate their orientations.Its three steps are local normal and curvature estimation, region growing, and orientation estimation from the completed point-cloud patch.Source: §3 · p. 6
3.1. Local surface normal and curvature estimation
- Each point’s local topography is represented by its K nearest neighbors; PCA-based least-squares plane fitting estimates normals, while covariance eigenvalues provide curvature.The neighborhood size k is user-chosen, and curvature is estimated without first creating a surface from the point cloud.Source: §3.1 · pp. 7–8
3.2. Region growing
- Growth starts from the unassigned point with minimum curvature, then adds neighboring points whose normals fall within the user-set deviation threshold.New seeds must also satisfy a transmission-error threshold relative to the initial seed’s normal, bringing the fracture’s overall occurrence into growth control; growth repeats until no new seed is selected.Source: §3.2 · pp. 8–9
- The normal-deviation threshold reflects local roughness, whereas the transmission-error threshold reflects overall fracture flatness; their useful values depend on geometry and weathering.A rough but flat fracture may require a larger normal-deviation threshold; a smooth but uneven weathered fracture may require a larger transmission-error threshold.Source: §3.2 · p. 9
3.3. Fracture orientation estimation
- After region growth, fracture orientation is estimated from the normal of a least-squares plane fitted to the extracted point-cloud patch.The normal can be converted to dip direction and dip, with the conversion depending on the point cloud’s geographic coordinate system.Source: §3.3 · p. 11
4. Results and discussion
The method’s fracture-orientation results were reported as comparable to or better than manual surveys, while also identifying small fracture faces that conventional measurements could not capture. Runtime rose steeply near one million points, and detected-plane counts varied with the configuration thresholds.
4.1. Details of results for a portion of the point cloud
- In a portion of the point cloud, most extracted fractures matched identifiable real fractures, and most major visible fractures were extracted.The results also report automatic extraction of the full extent of every individual fracture.Source: §4.1 · pp. 12–13
4.2. Results for entire point cloud and comparison with manual field survey results
- The method extracted many small fracture faces that conventional measurements could not obtain, potentially adding information about fractures to the survey.The authors identify small fractures in particular as a source of additional information.Source: §4.2 · p. 14
- The algorithm recovered major fracture-set clusters and was reported to locate them more accurately than the manual survey, yielding fracture data of equal or better quality.The comparison used pole-density plots for 65 manual orientations and 157 algorithm-estimated orientations.Source: §4.2 · pp. 15, 17, Figure 5 · p. 16
- The authors noted possible outlier clusters in the fracture sets and left their nature and removal for future study.Examples were visible in the algorithm’s pole-density plot.Source: Figure 5 · p. 16, §4.2 · p. 17
4.3. Performance and parameter configuration of the proposed algorithm
- For the tested point-cloud datasets, runtime rose steeply near one million points but remained acceptable; the authors judged the algorithm suitable for outcrop areas up to 70 m².For larger areas, they recommend increased computing power or modifying the algorithm.Source: §4.3 · p. 18, Figure 6 · p. 19
- The authors’ best-judged thresholds, θth = 6° and tth = 20°, were turning points: detected-plane counts declined quickly before them and more slowly afterward.Increasing either threshold reduced the number of detected planes; the authors suggest this analysis may help select a configuration.Source: §4.3 · p. 19, Figure 7 · p. 20
5. Conclusion
The region-growing method extracts individual outcrop fractures from 3D point clouds, estimates their orientations, and was reported to produce fracture data as good as or better than a manual survey. Its future development includes outlier removal and evaluation across rock types and weathering conditions.
5. Conclusion
- Local surface normals and curvature define fracture surfaces; thresholds for normal deviation and transmission error control region growth, and orientations are estimated for each extracted fracture.The transmission-error threshold considers the overall fracture occurrence to prevent unchecked growth.Source: §5 · p. 21
- The method automatically extracted the full extent of individual fractures and produced fracture data as good as or better than a manual field survey.Testing used a real outcrop point cloud and a manual survey of the same outcrop.Source: §5 · p. 21
- Future work should remove possible non-fracture outlier clusters, analyze fracture roughness alongside weathering, type, and orientation, and assess performance across rock types and weathering conditions.Source: §5 · p. 22