change-detection
Change analysis, once the observations are comparable. Not for cases whose blocker is comparability itself: mixed sensors, product levels or processing baselines to remote-sensing-analysis, undocumented vertical datums to point-cloud-lidar, multi-decade archive trends over large areas to google-earth-engine. Matching product level does not prove comparability. Otherwise invoke for what, where or how much changed: two-scene comparison, deforestation, urban growth, disaster damage, parcel-change audits, bi-temporal differencing, post-classification comparison, adjusted area, break detection in a series in hand (BFAST/LandTrendr/CCDC). Seasonal mismatch is this skill's own confounder; a documented datum with a stated accuracy budget is settled comparability. Keep both.
Muhammed Enes Duran · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add muend/geoai-skills --skill change-detection技能说明
name: change-detection description: >- Change analysis, once the observations are comparable. Not for cases whose blocker is comparability itself: mixed sensors, product levels or processing baselines to remote-sensing-analysis, undocumented vertical datums to point-cloud-lidar, multi-decade archive trends over large areas to google-earth-engine. Matching product level does not prove comparability. Otherwise invoke for what, where or how much changed: two-scene comparison, deforestation, urban growth, disaster damage, parcel-change audits, bi-temporal differencing, post-classification comparison, adjusted area, break detection in a series in hand (BFAST/LandTrendr/CCDC). Seasonal mismatch is this skill's own confounder; a documented datum with a stated accuracy budget is settled comparability. Keep both. license: MIT metadata: author: Muhammed Enes Duran
Change Detection & Spatio-temporal Analysis
Purpose: separate real surface change from the four great impostors — misregistration, radiometric drift, phenology, and classification error. Every method below exists to control one of them; skipping the controls produces confident maps of nothing.
Preconditions (where change detection is won or lost)
Preconditions 1 and 2 are checked here but established elsewhere. When either fails, the owning skill leads and this skill resumes once comparable observations exist. Precondition 3 is this skill's own problem and is never a reason to route away.
- Co-registration: sub-pixel alignment between dates (AROSICS or
manual tie-points). Half a pixel of shift creates edge-shaped phantom
change everywhere. Verify: flicker-compare crisp features. For elevation
surfaces or point clouds, vertical datum agreement, co-registration and
the vertical-accuracy budget belong to
point-cloud-lidar— a datum offset is not subsidence. - Radiometric consistency: same processing level (surface
reflectance), same sensor, same processing baseline. If any of the three
differ, this is a harmonization problem, not a thresholding one: hand it
to
remote-sensing-analysis(HLS for Landsat↔Sentinel-2, relative normalization with PIFs,BOA_ADD_OFFSETacross the Sentinel-2 2022 baseline change). - Same season / phenological stage for bi-temporal work — a May vs September pair "detects" summer. If season can't be matched, use composites or time-series methods instead.
- Cloud/shadow masks intersected across dates; analyze only mutually valid pixels and report that coverage %.
Method selection
| Situation | Method |
|---|---|
| Two dates, continuous "how much" | Index differencing (ΔNDVI, ΔNBR...) with statistical thresholding |
| Two dates, categorical "from-what-to-what" | Post-classification comparison (only with strong classifiers) |
| Two dates, multivariate robust | Change vector analysis (CVA); MAD/iMAD for sensor-robust detection |
| Dense stack, gradual + abrupt | Trend + break analysis (BFAST/LandTrendr/CCDC family; at archive scale → google-earth-engine) |
| Structure change (buildings) | DL bi-temporal segmentation (siamese U-Net) → geo-deep-learning |
| SAR pairs (clouds, disasters) | Log-ratio of calibrated backscatter + speckle handling |
| Vector vintages (parcels, buildings) | Geometry+attribute diff with tolerance (below) |
Thresholding — never eyeball it
Difference images need a defensible threshold: μ ± k·σ on the difference histogram (report k), Otsu when bimodal, or supervised thresholds calibrated on labeled change/no-change samples. Deliver the histogram with the chosen cut marked. Sensitivity: report changed-area at k-0.5 and k+0.5; if the story flips, the detection is fragile — say so.
Post-classification comparison (PCC) — handle with care
PCC error compounds: two 90%-accurate maps yield ≤ ~81% change accuracy, and biased errors create systematic false transitions. Rules:
- Use ONE classifier trained on both dates' imagery (same legend, same features) rather than two independent legacy maps.
- Build the full transition matrix (from-class → to-class areas), not just a change/no-change binary — impossible transitions (water→forest in 1 year) are your error detector.
- Apply a minimum mapping unit consistent across dates before differencing.
Time-series (dense stack) analysis
- Build a gap-filled, cloud-masked index stack (xarray, time dimension).
- Decompose trend + seasonality + breaks; per-pixel linear trends need significance testing (Mann-Kendall + Sen's slope for monotonic trends — and FDR correction across millions of pixels, or your "greening map" is noise).
- Label break DATES, not just presence — timing is usually the analytic payload (when did clearing start?).
- Validate detected breaks against known events (fires, construction permits, disaster dates) wherever records exist.
Vector change audit (two vintages of the same layer)
- Match features by stable ID if it exists; else spatial matching with IoU threshold (report it).
- Classify: added / removed / geometry-changed (area delta > tolerance) / attribute-changed. Tolerances absorb digitization jitter — 1-2 m for cadastre-grade, more for digitized-from-imagery.
- Sum area deltas by class and reconcile totals; unexplained residual = matching bugs.
Accuracy assessment (the deliverable's spine)
Change is rare, so random sampling wastes effort on stable pixels — use stratified sampling (strata: change/no-change or per-transition) with good-practice area estimation (Olofsson et al. protocol): report user's/producer's accuracy per stratum AND area estimates with confidence intervals adjusted for map error. A raw pixel count of the change map is a biased area estimate — always say the adjusted number.
Reporting template
## Change: <phenomenon>, <T1> → <T2 or period>
- Data: <sensor/level>, co-registration RMSE: <px>, valid overlap: <%>
- Method: <...> threshold/param
阅读完整来源 ↗