Changelog

All notable changes to RESIDUALS are documented here.

[Unreleased]

pyproject.toml for modern Python packaging (replaces bare requirements.txt as primary config).
Ruff linter/formatter configuration.
GitHub Actions CI pipeline (lint + test on Python 3.9/3.11/3.12).
.editorconfig for consistent editor settings.
CONTRIBUTING.md with development setup and method registration guide.
.env.example documenting the RESIDUALS_LIDAR_DIR environment variable.

LiDAR-dependent scripts (tile_county.py, trace_road.py, generate_licking_dem.py, generate_test_dem.py) now accept --lidar-dir CLI arg or RESIDUALS_LIDAR_DIR env var instead of hardcoded paths.

County grid tiling: blanket Licking County with ~200 tiles, each producing a 12-panel archaeological visualization (SVF, openness, RRIM, multi-scale TopHat). (tile_county.py)
Road corridor trace: extend Hopewell Road detection across the county with consensus heatmap and GeoJSON centerline. (trace_road.py)
Leaflet tile viewer for browsing county results in a web browser. (build_viewer.py, viewer.html)
Archaeological demo with side-by-side hillshade vs. residual comparisons for Great Hopewell Road and Great Circle Earthworks. (demo.py)
Best-of pipeline: 20 cluster representatives x 3 upsamplers = 60 combinations. (run_bestof.py)
166 pytest tests covering all methods, registries, known-answer correctness, and analysis.
Project refactor: extracted src/ library structure with registry-based decomposition and upsampling, analysis module, and utilities.

Redundancy analysis: SHA-256 checksums + statistical fingerprinting across all 39,731 combinations.
Identified 20 distinct method clusters and 3,345 exact duplicate groups.
Parallel fingerprint generator optimized for HDD I/O.
Rolling ball fix for large radii using downsampling approach.
Resumable fingerprinting with checkpoint support.
Prior art documentation (PRIOR_ART.md) with full method parameter spaces and results.

Initial release: 4-level differential framework for archaeological feature detection in LiDAR DEMs.
25 decomposition methods (Gaussian, bilateral, wavelet, morphological, tophat, polynomial, and 19 extended methods).
19 upsampling methods (bicubic, Lanczos, B-spline, FFT zero-pad, and 15 extended methods).
Exhaustive parameter exploration: 39,731 combinations generating 4.28 TB of prior art data.
Apache License 2.0.