Lunar Crater Detection Pipeline

Boundary detection of dispersal impact craters based on morphological characteristics, using lunar Digital Elevation Model (DEM) data.

Based on: Liu et al. (2017) — Boundary Detection of Dispersal Impact Craters Based on Morphological Characteristics Using Lunar Digital Elevation Model.

Pipeline Stages

1. D8 Flow Direction — Computes ArcGIS-style flow directions from DEM
2. Raw Crater Regions — Identifies depressions via priority-flood sink filling and D8 watershed expansion
3. Shape Classification — Classifies regions as strip / radial / suborbicular / square (§III-B)
4. Independent Craters — Filters false craters using posture ratio, rectangle factor, and sink validation (§III-C)
5. Affiliated Craters — Detects dispersal crater rims via profile analysis and inflection point detection (§III-D)

Code Files

• core.py — D8 flow direction computation
• preprocess.py — Sink filling (priority-flood) and depression core labeling
• depression.py — Depression identification, basin expansion, and morphological metrics
• false_crater_filter.py — Independent crater filtering (§III-C)
• affiliated.py — Affiliated dispersal crater detection with profile analysis (§III-D)
• cli.py — CLI entry point for the full pipeline
• io.py — Raster and GeoJSON I/O utilities
• validate.py — D8 flow direction validation
• directions.py — D8 neighbor encoding constants
• _numba.py — Optional Numba JIT acceleration

Usage

# Full pipeline (generates all outputs)
python -m code.cli <DEM.tif> --paper-shortcut

# D8 flow direction only
python -m code.cli <DEM.tif> <output_flow.tif>

---

Key Detection Thresholds (defaults)

Parameter	Default	CLI Flag	Purpose
Min crater depth	7.0 m	--crater-min-depth	Fill-depth threshold for a cell to be classified as a crater sink
Min region size	9 cells	--min-region-cells	Minimum watershed-expanded basin area
Max posture ratio	1.12	--posture-ratio	Max L/W ratio for shape acceptance (strip vs. radial)
Sink-fill method	priority-flood	--preprocess-method	Algorithm used for hydrological conditioning

Dataset Specifications

Input DEM

Property	Requirement
Format	GeoTIFF (.tif) — single-band, 32-bit float
Data type	Elevation in metres (float32 / float64)
CRS	Any projected CRS with metric units; pixel size is read directly from the raster transform
Nodata	Any sentinel value or NaN; auto-detected from raster metadata, overridable via --nodata
Band count	Single band (band 1 is read)
Minimum size	≥ 3 × 3 cells (smaller grids skip sink-filling)

Source: The pipeline was designed and validated against the Lunar Reconnaissance Orbiter Camera (LROC) NAC DEM tiles, as used in Liu et al. (2017). Any single-band elevation raster with consistent metric units will work.

📦 Datasets: bhargav247/DIP26_G01_Datasets on Hugging Face

File Resolution & Cell Size

• Cell size is derived from the affine transform (|a| for x, |e| for y).
• Can be overridden via --cellsize-x and --cellsize-y for rasters without a valid transform.
• D8 flow direction uses the true diagonal distance √(cellsize_x² + cellsize_y²) for slope comparison.

Output Files

File	Type	Description
<prefix>_flow.tif	GeoTIFF uint8	D8 flow direction (ArcGIS encoding: 1/2/4/8/16/32/64/128)
raw_crater_regions.tif	GeoTIFF int32	All watershed-expanded depression labels
crater_regions_strip.tif	GeoTIFF int32	Strip-classified regions
crater_regions_radial.tif	GeoTIFF int32	Radial-classified regions
crater_regions_suborbicular.tif	GeoTIFF int32	Suborbicular regions
crater_regions_square.tif	GeoTIFF int32	Square-classified regions
crater_labels_independent.tif	GeoTIFF int32	False-crater-filtered independent craters
crater_rims_independent.geojson	GeoJSON	Independent crater rim polygons with metrics
affiliated_analysis.geojson	GeoJSON	Profile lines, inflection points, hulls
crater_rims_affiliated.geojson	GeoJSON	Final affiliated crater rims