Platform / Captures & imagery

Captures, cataloged & delivered.

EO/SAR scene metadata - Sentinel-1-style - normalised into one canonical JSON product with a bounding-box footprint, machine-readable quality flags and a GeoJSON geometry, then cryptographically signed into an append-only chain of custody. A TRL-4 representative demonstrator: scene metadata plus bbox, not pixel or SAR processing.

InputsGeoTIFF, JPEG, PNG, provider scene + sidecar OutputsCanonical JSON product, GeoJSON footprint, JSON Schema

Stop writing a bespoke parser for every EO/SAR feed.

Every new imagery source arrives in its own native shape - a Sentinel SAFE package here, a provider scene plus sidecar there - so each integration becomes one-off glue code that rots, and format drift silently breaks pipelines downstream. Level Zero removes that: the EO/SAR lane ingests representative Sentinel-1-style scene metadata - scene ID, platform, acquisition mode, resolution and an acquisition bounding box - and normalises each one into a single canonical shape your tooling reads once: a validated canonical JSON product with a UTC acquisition timestamp, a GeoJSON Polygon footprint and machine-readable quality flags. It handles scene metadata plus a bounding box, not pixels or SAR processing.

Then it proves the record. Each canonical scene is cryptographically signed and hash-chained into an append-only chain of custody - enforced at the database layer, so a record can't be silently edited or deleted - and can be re-checked by anyone through the independent public verifier, so a bundle re-signed with a forger's own key is rejected as a forged or untrusted signature. Raw pixel handling and downstream raster science sit outside this TRL-4 demonstrator.

[ Why catalog & passthrough wins ]

Normalising to one canonical shape means the source of truth is always a signed, hash-chained record - fully auditable, with no hidden re-derivation. That focus keeps the layer small and dependable, running on a lean standard open-source stack and a geospatial database: a metadata normaliser, a canonical-product builder, a signing step and an append-only custody chain. Heavier radiometric and geometric processing is downstream roadmap, not part of this demonstrator.

Supported inputs

ClassFormat / packageNotes
EO sceneSentinel-1/2-style scene metadata + JSON sidecarAcquisition time, footprint bbox and platform read from the scene record.
SAR sceneSentinel-1-style scene metadata + bboxNormalised as scene metadata; acquisition mode and resolution stored, geometry as a bbox Polygon. No SAR/pixel processing.

Scene records are ingested as representative/synthetic inputs, tagged by how each was sourced - synthetic, replayed from file or pulled live; there is no live connector yet, so live pull is roadmap. Every input is normalised, then signed and chained before it enters the append-only custody chain.

Cataloging chain

Normalisation runs as small, declarative stages, and the pipeline is deterministic: the same inputs always produce the same signed outputs. A reproducible demo bundle and an automated test suite in CI keep it from drifting silently.

  1. Receive & hash - checksum the asset, store it immutably, record provider lineage.
  2. Read metadata - acquisition time, footprint bbox, platform, acquisition mode and resolution from the scene record.
  3. Validate - required fields and geometry checked against your schema; failures route to quarantine.
  4. Sign & chain - the canonical product is cryptographically signed and appended to the hash-chained chain of custody, with best-effort RFC-3161 trusted timestamps.
  5. Expose - the signed product is reachable over the documented REST API and the public verifier. (A generic signed outbound-webhooks module exists at the platform level but is not yet wired to emit Level Zero products.)

Each stage emits structured telemetry: input hash, extractor version, parameters, timing, and a signed receipt. Nothing about a capture is opaque.

Outputs

Canonical JSON

One canonical product with a UTC timestamp, source block, GeoJSON geometry, quality flags and a 0.0-1.0 confidence - validated against a published JSON Schema.

Footprint

The capture outline as a GeoJSON feature, ready for map views and spatial search.

Signed provenance

An Ed25519 signature and hash-chained custody record, verifiable through the independent public verifier pinned to a published trust anchor.

JSON Schema & crosswalk

A machine-readable JSON Schema for the canonical product, plus an OGC SensorThings 1.1 alignment crosswalk (a field-mapping table, not a certified server).

Quality & QA

Every canonical product carries machine-readable quality flags (for example flagged as spoofable on live-tracking inputs) and a 0.0-1.0 confidence, and is validated against the published JSON Schema before it is signed and chained. The whole pipeline ships a reproducible demo you can run yourself, backed by an automated test suite in CI.

For audit and reproducibility, every canonical record is cryptographically signed and hash-chained into an append-only chain of custody that cannot be silently rewritten.

Delivery

Canonical products are reachable via the documented REST API - product and feed endpoints, a GeoJSON corroboration feed, the schema, a public verifier and its published trust anchor. There is no published SDK yet; a generic signed outbound-webhooks module exists at the platform level but is not wired to Level Zero. See the REST API docs.

[ Ready to integrate ]