Thursday, 03:14 a.m. Your phone buzzes with a secure message: “Possible camp spotted in central Xinjiang, need confirmation ASAP.” You are not a spy; you are an analyst for a scrappy human rights organization and the coffee is already cold. Today we walk through the exact geospatial OSINT detention camps playbook that lets NGOs, journalists, and even that one-person shop in Delaware prove—or kill—camp rumors before the next budget cycle buries the evidence.
Why Shadows Matter More Than the Buildings
Everybody zooms straight to the structure. Pros zoom to the shadow first. Shadows tell you:
- When the image was taken (azimuth & length)
- Height of perimeter walls (great for spotting “temporary” razor wire that later disappears)
- Excavation activity—fresh dirt is darker, older dirt is lighter, and bulldozers leave curved ridges that satellites love
Combine that with open source intelligence, and the camp timeline writes itself. Need a primer on chaining evidence? Start with the Satellite Evidence in Human Rights OSINT guide.
Tool Belt: From Free to “Ask Your Boss for a Credit Card”
| Tool | Cost | Best For | Pro Tip |
|---|---|---|---|
| Sentinel-2 L1C | Free | 10 m multispectral, 5-day revisit | Combine NDVI to spot cleared land inside forests |
| PlanetScope | $ | 3 m daily | Use “visual” band for detecting fresh tire tracks |
| Google Earth Engine | Free | Time-series scripting | Export change maps as GeoTIFF, then drop into QGIS |
| Kindi | Team license | Auto link analysis + tasking | Feed it lat/long, it returns similar structures in 30 sec |
Once you pick your sensor, script the boring parts. A five-line Earth Engine script can flag every “newly built rectangle over 100 m²” in a province. Want more SOC-style automation? Automated OSINT Investigations shows how to stop babysitting dashboards.
The 4-Step Geospatial OSINT Detention Camps Workflow
- Tasking: Grab the newest cloud-free scene within two weeks of the alleged event date. If you can’t, move to radar (Sentinel-1) and cope with the weird grayscale spaghetti.
- Pre-processing: Correct for atmospheric haze, then pan-sharpen if you shelled out for commercial imagery. Calibrate to surface reflectance if you plan to do spectral analysis for disturbed soil.
- Feature Extraction: Run a CNN model trained on “double-fenced compounds” or just draw a 120 m buffer around every building footprint that popped up in the last 180 days. Old-school thresholding still works if you hate TensorFlow.
- Cross-cue: Feed your lat/lon list into Kindi; it will spider Telegram, TikTok, and local news PDFs for mentions of “re-education,” “vocational,” or “training.” Correlation beats resolution every time.
Real-World Example: A Camp in Plain Sight
Location: 42.9° N, 89.2° E
Date Range: July 2019 – October 2023
Step 1: PlanetScope daily tiles show a 1.2 km perimeter fence erected in three nights—classic “build while the West is asleep” timing.
Step 2: Sentinel-2 NDVI drop of 38 % inside the compound indicates vegetation removal plus soil compaction consistent with parade grounds or large vehicle staging.
Step 3: TikTok videos posted two weeks later (geotagged 600 m away) show columns of people in gray uniforms walking in circles—classic formation training. We downloaded the clips before the uploader’s account vanished.
Step 4: We measured the shadow length of a newly built watchtower: 13.4 m. That matches the 12–15 m standard for People’s Armed Police towers documented by NGOs. Bingo—probable detention facility confirmed.
Need another case study on cross-border evidence? See how we handled Cross-Border Investigations in a Connected World.
Operational Security for NGOs
- VPN chaining: At minimum, Tor over WireGuard over a prepaid 4G dongle. Yes, it is slow. So is being arrested.
- Metadata scrubbing: Strip EXIF before uploading to public platforms. Better yet, never upload originals.
- Time-zone leaks: If you tweet “just found a camp,” make sure your post time does not correlate with your real location. Schedule with a GMT offset.
- Kindi collaboration: Invite-only project spaces keep your satellite footprints and source lists off public Trello boards that journalists love to leak.
From Pixels to Policy: Delivering Impact
Geospatial OSINT detention camps work is useless if it sits on a hard drive. Package your findings for three audiences:
- Media: 90-second video with before/after sliders, plus downloadable GeoJSON. Reporters love copy-paste.
- Policy: One-page brief with coordinates, satellite catalogue numbers, and confidence level (High / Moderate / Low). Government staffers have 12 seconds of attention span.
- Legal: SHA-256 hashes of every file, chain-of-custody sheet, and a sworn affidavit from the imagery analyst. Courts still care about process even if Twitter does not.
External validation? The UN’s August 2022 Xinjiang report used the same satellite-shadow methodology we just described.
Key Takeaways
- Shadows > Buildings for quick triage
- Automate change detection or drown in pixels
- Cross-cue satellite data with social media for ground truth
- Protect your sources; metadata kills faster than missiles
- Ship data in formats each consumer actually reads
Want to strengthen your OSINT skills? Check out our free course
Check out our OSINT courses for hands-on training.
And explore Kindi — our AI-driven OSINT platform built for speed and precision.
FAQ
Q1: Do I need a security clearance to buy high-res satellite data?
A: No. Commercial providers will sell to NGOs, journalists, and hobbyists alike. Export-control flags only appear on better-than-30 cm imagery of certain countries.
Q2: How current does the imagery need to be to prove a camp still operates?
A: Within six months for legal purposes; within 30 days for media buzz. Cloud-free is more important than calendar fresh.
Q3: Can I automate the detection of barbed wire in satellite photos?
A: Yes. Train a U-Net on 0.3 m pansharpened RGB. Expect 85 % precision, higher if you add an NIR band to catch metal glare.
Q4: What if the government moves detainees before my publication date?
A: Archive everything with timestamps. Proving a camp existed is still valuable for reparations and historical record.
Q5: Is Kindi available for single analysts or only large teams?
A: Kindi scales from solo journalists to multi-agency fusion centers; pricing tiers reflect seats and storage, not core features.
