skysatprep 0.2.0

Radiometric prep for SkySat L1A panchromatic stereo with RPC preservation and optional RPC-quicklooks

SkySatPrep

Authors: Aram Fathian; Dan Shugar

Affiliation: Department of Earth, Energy, and Environment; Water, Sediment, Hazards, and Earth-surface Dynamics (waterSHED) Lab; University of Calgary

License: MIT

Radiometric preprocessing for SkySat Basic L1A panchromatic stereo imagery. The tool performs robust radiometric correction (percentile stretch + optional CLAHE + conservative shadow/highlight tone curve), preserves and embeds RPC metadata, and can generate RPC‑based quicklook orthos (via GDAL) to aid QA.

Input data: single‑band 16‑bit GeoTIFF frames matching the SkySat L1A PAN naming convention (*_basic_l1a_panchromatic_dn.tif) with sidecar metadata (e.g., .RPB, _RPC.TXT, .json, .imd, .xml). The tool is optimized for stereo pairs but also works on individual L1A PAN frames.

Output data: radiometrically corrected 16‑bit GeoTIFF with embedded RPC and optional internal pyramids. Quicklook orthos (optional) are generated using gdalwarp -rpc and a user‑supplied DEM.

• Robust percentile stretch (--pmin/--pmax)
• Optional CLAHE (local contrast) if OpenCV is available
• Conservative tone curve to lift shadows while protecting highlights
• Copies SkySat sidecar files and embeds RPC in the output TIFF
• Optional gdaladdo pyramids
• Optional gdalwarp -rpc quicklooks with RPC_DEM=...

Install

You need GDAL (>=3.4) available to Python and the gdalwarp/gdaladdo CLI on PATH.

# create env (recommended)
python -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate

# install from source
pip install -U pip
pip install -e .[dev]     # or: pip install -e .

# If you want CLAHE:
pip install opencv-python

Usage

skysatprep   --pair1_src /path/to/Pair1/SkySatScene   --pair2_src /path/to/Pair2/SkySatScene   --pair1_out /path/to/Pair1/radprep   --pair2_out /path/to/Pair2/radprep   --pmin 1 --pmax 99   --clahe 3.0 --tiles 8   --shadow_boost 0.20 --highlight_comp 0.10   --pyramids   --quicklook --rm_quicklook   --dem /path/to/ellipsoidal_dem.tif   --t_srs EPSG:32608 --ql_res 1.5

Notes

• Input files must match *_basic_l1a_panchromatic_dn.tif.
• Sidecars found next to each input (e.g., .RPB, _RPC.TXT, .json, .imd, .xml) are copied to the output directory and RPC is embedded into the processed TIFF.
• Quicklooks use gdalwarp -rpc with -wo RPC_DEM=<DEM> for approximate ortho + resampling; they are not a substitute for rigorous photogrammetry.

Metashape tips

• Import the processed 16-bit panchro TIFFs; Metashape will see the embedded RPC.
• Use your usual SfM settings for SkySat stereo (align, build DSM/mesh) and any sensor-specific tweaks.

Reproducibility

• Default stretch is --pmin 1 --pmax 99. Pin these and the tone-curve parameters in your paper’s methods to ensure consistent output.
• Consider committing a small config and logging the full CLI (see your shell history).

License

MIT (see LICENSE).

Data handling & limits

• Expects single‑band panchromatic L1A frames (PAN). Multispectral products are not processed by this tool.
• File discovery defaults to the SkySat PAN L1A naming pattern. Adjust EXT_TIF_RE in core.py if your filenames differ.
• Quicklook orthos are for QA/visualization; they are not a substitute for rigorous photogrammetric processing.
• For best quicklooks, use an ellipsoidal‑height DEM (or a DEM consistent with the scene’s RPC model).

How to cite

Fathian, A., Shugar, D. (2025). SkySatPrep (v0.1.0) [Software]. Zenodo. https://doi.org/10.5281/ZENODO.17156601