Run bbox / pixel detection

bbox_change_detector.py runs a pixel-level raster workflow over a bbox or vector AOI.

AOI_SOURCE = "bbox"   # or "vector"
AOI_BBOX = {"west": 6.9563, "east": 6.9868, "south": 61.2817, "north": 61.3012}
AOI_VECTOR_PATH = None

Common constants: REFERENCE_START, MONITOR_START, MONITOR_END, Z_THRESHOLD, PERSISTENCE_N, MIN_HISTORY_SCENES, DURABILITY_N, MIN_MAX_Z, DROP_WINTER, and WINTER_MONTHS.

Outputs under results/pixel_detection/ include flagged_mask_raw.tif, flagged_mask.tif, change_date.tif, max_z.tif, change_doy.tif, and run_metadata.json. Use vectorize_results.py to create detected_changes.gpkg.

Exploratory, not final

Whole-AOI scans are useful for diagnostics but can produce many false positives. Operational work should start from candidate polygons.