FREE TRIAL

Automated Region Detection in ShapeMetriX: The Fastest Path from 3D Model to Structural Analysis

Structural mapping of rock masses remains a time-intensive, analyst-dependent task in geotechnical investigations. Identifying planar discontinuity surfaces across complex 3D geometry requires both experienced judgment and significant manual effort; effort that scales poorly as scan resolution increases and project scope widens.

ShapeMetriX, 3GSM’s photogrammetric rock mass characterization platform, disrupts this paradigm through its Automated Region Detection tool. It scans the full 3D mesh geometry, autonomously identifies coplanar surface patches, and renders them color-coded in the viewport – all before a single measurement is manually placed. The result is a dramatically faster path from raw scan data to defensible structural interpretation.

The Region Detection tool is not simply a convenience feature. For high-resolution 3D models spanning large rock face extents, it is the only practically viable mechanism for capturing the full statistical population of discontinuity orientations at scale. This article unpacks how the algorithm works, how to tune it for different rock mass conditions, and how the downstream tools integrate into a complete geotechnical analysis pipeline.

Figure 1. Fully automated mapping of areas on a 3D model, resulting in several hundred measurements, with the corresponding stereographic projection of mapped planes.

The Region Detection Algorithm

At its core, Region Detection searches for planar patches (referred to as regions) directly triangulated on the 3D model mesh. The algorithm groups surface elements that meet a user-defined planarity criterion into distinct regions, assigns each region a unique identifier, and computes the plane orientation, including dip direction and dip angle.

The detection process is resolution-dependent and works on the geometry as displayed in the 3D viewer. The quality of the geometry directly affects how the algorithm identifies planar patches: higher-quality geometry allows finer, more detailed regions to be detected, while lower-quality geometry may merge features into larger, less-defined regions. Adjustments to geometry quality, which can be configured independently, influence both the number and size of detected regions and the clarity of fine structural details, ensuring that the analysis captures the level of detail appropriate for the model and investigation scale.

Workflow

The Region Detection workflow follows a structured sequence that converts 3D surface geometry into discrete structural measurements:

  • Detection of regions based on surface geometry and planarity, using mesh resolution and quality parameters.
  • Visualization of detected regions on the 3D model with colour coding, along with orientation poles plotted in the stereonet.
  • Refinement of the detected dataset using geometric filters.
  • Application of orientation filters to isolate specific structural sets.
  • Clustering of regions into candidate Structure Sets, with the ability to review and organize clusters.
  • Conversion of filtered and clustered regions into Areas, generating discrete, measurement-ready structural data with defined boundaries and properties.

Coplanarity: The Primary Detection Parameter

Coplanarity is the single most consequential parameter in the Region Detection workflow. It controls how strictly a surface patch must conform to a plane in order to be accepted as a region. Lower settings allow rougher, less planar surfaces to be grouped, resulting in regions that include more geometrically irregular areas. Higher settings detect only surfaces that are closely coplanar, producing smaller, cleaner regions while excluding any surfaces that fail to meet the threshold.

The practical effect of this parameter is best understood visually. At low coplanarity, irregular joint surfaces are merged into broad, colour-coded patches. At high coplanarity, the same face is partitioned into smaller, more strictly planar patches, and non-planar areas are excluded from any region.

Figure 2. Influence of increasing coplanarity: low (left), intermediate (middle), high (right)

There is no universally correct coplanarity setting. The appropriate value depends on the in-situ roughness of the discontinuity surfaces and the intended analytical purpose. For blocky, well-jointed rock masses with planar joint surfaces, a higher coplanarity setting is appropriate. For irregular, rough, or undulating surfaces (typical of tension fractures or foliated surfaces), a lower setting may be necessary to achieve meaningful region detection coverage.

Analysis Zones: Spatially Constraining Detection

Region Detection does not have to be applied to the entire 3D model. Engineers can define closed polygons directly on the 3D model surface that either confine detection to a specified area (Analysis Zones) or exclude defined areas from detection (Exclusion Zones). Multiple zones of both types can be active simultaneously.

This capability is more important than it first appears. Rock mass analyses rarely treat an entire scan as a single homogeneous domain. Common use cases include:

  • Characterizing different structural domains independently within the same scan. For example, separating a hanging wall, footwall, and crown in an open pit bench, or distinguishing fresh rock from a blast-damaged or heavily weathered zone.
  • Confining detection to a specific bench face, tunnel wall segment, or mapped exposure of interest within a larger scan.
  • Excluding surface irregularities, overbreak zones, or structurally disturbed areas that would contaminate the primary structural dataset.
Figure 3: Region detection in user-outlined zones: Zone 1, Zone 2, and an Exclusion Zone are all active simultaneously

Visualization Controls

Following region detection, results are displayed on the 3D model using configurable color coding that simultaneously updates the stereonet pole distribution.

  • Geotechnical (mirrored colours): Colour assignment mirrors conventional geotechnical stereonet representations, grouping orientation families into visually consistent chromatic bands.
  • Inclination (continuous radial, blue to red): Encodes a continuous colour gradient across the full inclination spectrum. Steep structures render red; shallow structures render blue. Effective for rapid visual assessment of kinematic risk across an excavation surface.
  • Inclination discrete (discrete radial, blue to red): The same encoding as Inclination but rendered in discrete colour steps rather than a continuous gradient. Useful where clear visual boundaries between inclination classes are preferred over a smooth ramp.
Figure 4. Color coding: Geotechnical (top) vs. Inclination (bottom)

Filter Options

Post-detection filters refine the working dataset without rerunning detection, updating the 3D model and stereonet in real time.

Geometric filters control the size and planarity of regions. The diameter filter limits regions by their extent, affecting both very small and oversized features, while coplanarity deviation measures how closely a region conforms to a best-fit plane, allowing selection of the most planar patches independently of the detection-time coplanarity setting.

Orientation filters isolate regions based on structural attitude, using either a cone-shaped directional window or a range of dip direction and dip angle values, with inclusive or exclusive selection. Together, these filters allow precise control over which regions are included in the working dataset.

Figure 5. Orientation filter: Cone mode (left) and Range mode (right), with corresponding stereonet cone.

Joint Set Clustering

Joint Set Clustering within the Region Detection tool automatically groups detected regions into Structure Sets based on orientation similarity. The clustering engine is a fuzzy K-means algorithm applied to the pole distribution.

The rationale for fixing these parameters within Region Detection is workflow efficiency: the intent is to rapidly identify candidate Structure Sets from the raw region population. Individual clusters can be managed and edited before Area generation. Selecting or deselecting clusters immediately updates the 3D model and stereonet, giving users full control over which clusters are included in the analysis.

Figure 6. Clustering in the Region Detection tool: before clustering (left) and after a 5-cluster solution (right)

Area Generation

Area generation is the terminal step of the Region Detection workflow and the point at which detected regions are converted into discrete, measurement-grade structural data.

Each generated Area has a closed polygonal boundary that follows the discontinuity surface and carries the following properties per measurement: size (surface area), dip direction, and dip angle. These properties are immediately available in the Structure Set list and are compatible with the full suite of Analyst analysis tools, including Stereonet Analysis, Trace Map Analysis, and advanced Clustering tools.

The area generation step respects all active filters and cluster selections. Only regions that pass the active geometric and orientation filters, and belong to the selected clusters, contribute to the final measurement population.

Summary

The Region Detection workflow in ShapeMetriX is a revolutionary tool that redefines how structural analysis is performed. By consolidating surface selection, orientation measurement, set classification, and measurement generation into a single, intelligent workflow, it transforms a multi-stage process into a fast, automated, and highly precise operation. Its combination of detection, filtering, and clustering delivers comprehensive, measurement-ready datasets that capture the full geometric complexity of rock masses, even across large, high-resolution models. This ensures a statistically defensible interpretation.

Region Detection transforms structural analysis from a subjective, time-consuming task into a reproducible, statistically robust workflow, setting a new standard for efficiency, accuracy, and insight in geotechnical investigations.