## [2.0.0] - 2026-06-04

### Added — Export Pipeline (LiDAR Relief — Export group)

- **Cloud-Optimized GeoTIFF (COG) Export**: Convert any algorithm
output to a
  cloud-optimized GeoTIFF with internal tiling, overviews, and
DEFLATE/LZW/ZSTD
  compression. Optionally generates an interactive MapLibre GL JS web
viewer
  with opacity controls, coordinate display, dark/light themes, and
share-link
  copying. Suitable for direct upload to GitHub Pages or any static
host.
  (`export/cog_exporter.py`, `export/web_viewer.py`; algorithm:
`cog_export`)

- **Field Survey Export**: Package relief rasters and anomaly
detection points
  into a GeoPackage with structured archaeological schema (anomaly_id,
  detection_method, confidence, feature_type, field_status) plus a
QGIS project
  file that opens directly in QField/Mergin Maps for mobile
ground-truthing.
  (`export/field_packager.py`; algorithm: `field_survey_export`)

- **PDF Report Generator**: Generate CIfA-compliant PDF reports with
title page,
  full algorithm parameter documentation, input DEM metadata (CRS,
resolution,
  extent), band statistics with percentiles, histogram chart, and
certification
  section. Uses ReportLab (pure Python, OSGeo4W-safe).
  (`export/report_generator.py`; algorithm: `pdf_report`)

- **Visualization Recipes**: Import/export all algorithm parameters as
  shareable JSON files with versioned schema, validation, type
checking, and
  metadata fields (name, author, tags, landscape type). Enables
community
  sharing of optimized parameter presets beyond the 4 built-in
landscape
  presets. (`recipes/__init__.py`; algorithms: `recipe_export`,
`recipe_import`)

### Added — Point Cloud Processing (LiDAR Relief — Point Cloud group)

- **CSF Ground Filter (LAS/LAZ → DEM)**: Archaeology-tuned ground
extraction
  using the Cloth Simulation Filter (CSF) with 4 presets: Archaeology
Fine
  (maximum micro-relief preservation), Archaeology Standard
(balanced),
  Forested (aggressive canopy), and Urban. Supports laspy and PDAL
readers.
  (`point_cloud/csf_filter.py`; algorithm: `csf_ground_filter`)

- **PDAL Ground Classification Pipelines**: JSON-based pipeline
builder for
  Progressive Morphological Filter (PMF) ground classification with
  archaeology-optimized parameters. Presets for fine, standard, and
forested
  terrain, plus statistical outlier removal.
  (`point_cloud/pdal_pipeline.py`; 4 pipeline presets)

### Added — Multi-temporal Analysis (LiDAR Relief — Temporal group)

- **Multi-temporal Change Detection**: Compute a probabilistic DEM of
  Difference (DoD) between two temporally separated DEMs with
propagated
  RMSE-based Level of Detection (LoD) masking. Outputs include signed
DoD
  raster, significance mask (erosion/deposition), and cut/fill volume
report.
  (`temporal/dem_difference.py`; algorithm: `temporal_difference`)

### Added — Multi-Sensor Fusion (LiDAR Relief — Fusion group)

- **Multi-Sensor Fusion**: Co-register Sentinel-2 multispectral bands
with
  LiDAR relief and apply blend recipes combining topographic and
spectral
  information. Four recipes: Terrain+CIR (SVF + Colour Infrared), Crop
Mark
  Enhancement (Local Dominance + true colour), Erosion Risk (Slope +
SWIR),
  and Bare Earth Composite (SLRM + SWIR). Blend modes: luminance
overlay,
  overlay, multiply, screen.
  (`fusion/sentinel_fusion.py`; algorithm: `multi_sensor_fusion`)

### Added — AI/ML Inference (LiDAR Relief — AI/ML group)

- **AI Feature Detection**: Load user-provided ONNX models for object
  detection (YOLO) or semantic segmentation (U-Net) on plugin
visualizations.
  Features tiled processing for large rasters, Non-Maximum
Suppression,
  confidence filtering, and GeoPackage export of bounding box
detections.
  Pure NumPy preprocessing — no OpenCV dependency at inference time.
  (`ml/detector.py`; algorithm: `ai_feature_detection`)

### Added — GPU Acceleration

- **CuPy Compute Backend**: Transparent GPU acceleration for
computationally
  intensive horizon-scanning algorithms (SVF, Openness). Automatically
detects
  CUDA availability and dispatches to CuPy with graceful NumPy
fallback. Uses
  the trigonometric identity `sin(arctan(dz/d)) = dz / sqrt(dz² + d²)`
for GPU-
  optimized horizon scanning. (`gpu/compute_backend.py`)

### Fixed
- Fixed missing `compute_mstp` wrapper function in `core/mstp.py` that
caused an
  `ImportError` when running the e4MSTP algorithm via Batch mode.

None