ZeroTrace OSINT
Geolocation Clue Extractor
OCR, language and script detection, license-plate format matching, and signage colour analysis — for image-geolocation challenges.
The geolocation clue extractor is the moment-of-truth tool for image-geolocation challenges. Where most geo-clue tools make you type the clues you spotted, this tool finds them.
OCR-based text extraction, language and script detection, license-plate format matching, and signage colour analysis combine into a ranked list of country candidates with rationale per row.
What you get
For any image:
| Section | What it surfaces |
|---|---|
| OCR text dump | Every readable string the OCR found, classified by type |
| Language detection | Probable language(s) of any text found |
| Script detection | Latin / Cyrillic / Arabic / CJK / Devanagari / Greek / Hebrew — even when OCR fails |
| License-plate matches | Plate format patterns matched against the per-country catalog |
| Signage colour analysis | Dominant signage colour bands matched against per-country sign conventions |
| Driving-side hint | Right-hand drive / left-hand drive (when vehicles are visible enough for plate position) |
| Country candidates | Ranked list with per-country rationale |
OCR text classification
Every readable string the OCR finds is classified into:
| Type | Example |
|---|---|
| Phone number | International / national format detection |
| Standard email patterns | |
| Street name | Patterns like Str., Strasse, Avenue, Calle, улица, 通り |
| Business name | Capitalised multi-word phrases |
| License plate | Per-country format patterns |
| Other text | Everything else |
Classification is heuristic. False positives happen, particularly on low-quality images. The tool surfaces the raw OCR alongside the classified version so you can sanity-check.
Language and script
A purely-text-based language detector runs over the OCR output. For images with little text, the script detection is the more reliable signal:
- Latin script — Western Europe, Americas, much of Africa, Vietnam, parts of South-East Asia.
- Cyrillic — Russia, Belarus, Ukraine, Bulgaria, Serbia, Mongolia.
- Arabic — Middle East, North Africa.
- CJK — China, Japan, Korea, Hong Kong, Taiwan.
- Devanagari — North and Central India, Nepal.
- Greek — Greece, Cyprus.
- Hebrew — Israel.
A combination — Latin + CJK on the same sign — narrows even further (often a tourist area, an international business district, a transport hub).
License-plate format matching
The tool ships a curated catalog of license-plate formats per country. For example:
| Pattern | Country |
|---|---|
AB CD 1234 (yellow + EU stars) | EU member state |
ABC-1234 | Many US states (state-specific colour patterns differentiate) |
123 RU 77 | Russia (region code at the right) |
1-1234 (small white plate) | Common Japanese format |
A match — even partial — tells you the country of registration. If the plate's country of issue differs from the OCR'd text language, that itself is signal (a foreign-registered vehicle in a non-foreign-language environment is typically a tourist or a long-haul vehicle).
Signage colour analysis
The tool samples the dominant colours of any text-bearing rectangular regions in the image and matches against per-country signage conventions:
| Colour pattern | Common in |
|---|---|
| Red + white | Generic warning signs (most countries) |
| Blue | European motorway directional signs |
| Green | US highway directional signs |
| Yellow + black | German Autobahn warning signs (and many others) |
| Brown | Tourism / heritage signs (many countries) |
Sign colour is a coarse signal but cheap to compute. It rules out broad regions even when no specific clue is present.
Country candidates
The tool combines all the per-clue signals into a ranked list of country candidates, with the rationale shown per row:
Germany — high confidence Rationale: German-language street name detected (
Hauptstr.); EU yellow license plate format; blue motorway sign; Latin script.
Austria — medium confidence Rationale: German-language ambiguous between DE / AT / CH; sign style consistent with Austria; license plate format consistent.
Switzerland — low confidence Rationale: German-language consistent; license plate format does not match.
A typical strong-clue image narrows to one country with high confidence. A weak-clue image presents three to five candidates that you can then rule out further with aerial comparator checks.
The geo clue tool is a starting point, not an answer. It surfaces clues that narrow the search; it does not "solve" a geolocation. The next steps are usually reverse-image search, aerial verification, or talking to a regional contact who knows the area.
Pivots
| Click on... | Pivot to |
|---|---|
| Country candidate | Aerial comparator (centred on the country's capital initially) |
| Phone number from OCR | Phone lookup |
| Email from OCR | Email analyzer |
| Business name | (no pivot — copy and search externally) |
| License plate | (no pivot — copy and search externally) |
Sources
- OCR runs locally with a bundled model.
- Language detection runs locally with bundled language profiles.
- Script detection is rule-based, no external source.
- License-plate format catalog and signage colour catalog are bundled.
- Optional Wikidata lookup for "languages spoken in" cross-reference.
The OCR model is bundled with the application — the tool works fully offline.