svr-roughness

svr-roughness is a reusable Python library for SurfInspect-compatible surface roughness analysis from industrial scan outputs. Use svr_roughness as the consistent Python import name.

The core API is file-type agnostic: roughness is computed from an Nx3 NumPy-like XYZ point array in millimeters. Svr is calculated from signed distances to a smoothed, triangulated reference surface and a KD-tree variogram modeled on SurfInspect. The regular grid remains available for Sa/Sq and visualization. Reference mesh resolution and smoothing default to 0.30 mm, matching the supplied SurfInspect A2 reference closely. At most 10,000 variogram points are used by default to bound memory on industrial scans; set max_svr_points higher when full point coverage is required.

config = RoughnessConfig(
    mesh_resolution_mm=0.30,
    mesh_smoothing_mm=0.30,
    svr_points=10,
    svr_span_mm=0.50,
)

Install

From this repository:

python -m pip install ./svr-roughness

For deployment from inside SurfaceRoughnessPi:

python -m pip install ../svr-roughness

Python API

Use analyze_points() when your scanner or upstream software already gives you XYZ points:

import numpy as np
from svr_roughness import analyze_points

points_xyz_mm = np.asarray(points)  # shape (N, 3), columns x/y/z, units mm
result = analyze_points(points_xyz_mm, grid_mm=0.30, short_cutoff_mm=0.6, long_cutoff_mm=8.0)

print(result.sa_um, result.sq_um, result.svr_um)

Use analyze_file() as a convenience adapter for supported files:

from svr_roughness import RoughnessConfig, analyze_file

config = RoughnessConfig(grid_mm=0.30, short_cutoff_mm=0.6, long_cutoff_mm=8.0)
result = analyze_file("scan.ply", config=config)

Supported file loaders (all are converted to an Nx3 NumPy array):

• ASCII and binary little-endian .ply point clouds
• ASCII and binary .pcd point clouds
• ASCII and binary .stl mesh vertices
• .obj vertex meshes
• comma-, tab-, or whitespace-delimited .csv, .tsv, .xyz, and .txt
• .npy and .npz NumPy arrays

STL files are converted to their unique mesh vertices before analysis. For production mesh metrology, prefer scanner point clouds or add controlled surface sampling before calling analyze_points().

Result Output

result.save_grid_npz("output/roughness/latest_grid.npz")
result.save_metrics_json("output/roughness/latest_metrics.json")

save_metrics_json() writes:

• sa_um
• sq_um
• svr_um
• point counts, grid dimensions, grid coverage, and filter cutoffs

save_grid_npz() writes:

• grid_raw
• grid_filled
• grid_filtered
• valid_raw
• valid_filled
• grid_origin
• plane basis arrays

Command line

The package also includes a no-code command for scanner integrations:

svr-roughness scan.ply --grid-mm 0.30 --metrics-out metrics.json --grid-out grid.npz

Coordinates are assumed to be millimeters, and metrics are reported in micrometers. Unit conversion should happen before calling the library.