Kraken API (Imagery and Analyses)

Legends say that the Kraken was a giant sea creature that was bigger than the eye could see. Sometimes, one wants to get an overview of a bigger area than Ragnar API can provide. Borrowing its name from the Kraken, the sea monster, the Kraken API was born to serve this very need.

The API interfaces imagery and analyses through tiled web map interface.

Warning

https://spaceknow-kraken.appspot.com/ is deprecated and should not be used anymore.

For more information about algorithms available for Kraken see algorithms .

Map and Tiling

Kraken map is Earth surface between -180 to +180 of longitude and -85.051129 to +85.051129 of latitude projected to a square with Web Mercator projection. The map is divided into multiple zoom levels each consisting of a grid of tiles.

A tile is unambiguously identified by \((z, x, y)\) coordinates, where \(z\) represents zoom level, \(x\) represents a horizontal coordinate and \(y\) represents a vertical coordinate. The top-left tile has coordinates \((z, 0, 0)\).

A zoom level is a number that indicates the level of tile’s fine-detail. The lowest zoom level, at which the whole map is rendered to a single tile, is 0. With each successive zoom level, the total number of map tiles quadruples. The zoom level \(z >= 0\) is a grid with \(2^z \times 2^z\) tiles.

A tile is physically represented by one or many files representing area given by the tile boundaries. The files may consist of PNG and other images with dimensions of 256x256 pixels, a GeoJSON or other types of raster and vector data. See Kraken Pairwise API.

Glossary

  • Kraken map – an individual image or analyses identified with a unique ID.
  • Map type – type of a Kraken map. For example WRUNC or IMAGERY.
  • Grid file – a map is divided into tiles. For every tile (i.e. some zoom, x, y coordinates) there is one or more available grid files (e.g. truecolor.png or detections.geojson).

Release The Kraken!

Kraken release endpoints serve as a preparation and/or re-computation of a map. All release endpoints generate a map ID which can be later used to access the map. A map ID expires in one year.

Generic Kraken Release Endpoints

Kraken API can generate various map types. One can generate a map for a GeoJSON. Release endpoints have uniform URL paths in format: /kraken/release/<map_type>/geojson/{initiate|retrieve}.

Needed Permissions:

  • kraken.release
  • Also a set of permissions in the form <prefix>.<provider>.<dataset> is needed. <prefix> is replaced with algorithms.<name-of-algorithm> or with imagery.images.
POST /kraken/release/{map_type}/geojson/initiate
Request JSON Object:
 
  • sceneId (string) – required; ID of a scene from which the map should be generated. See Ragnar API (Search and Get Imagery).
  • extent (object) – required; area of interest which must intersect with scene’s footprint, see Extent. The area must be allocated using Credits API if the imagery is high-resolution.
  • subGeometry (object) – optional; Subset of extent of interest. If provided the response contains subGeometryId which can later be used in Kraken Grid endpoint to crop result of analysis. See Release with subGeometry.

Example request:

{
    "sceneId": "abc",
    "extent": {
        "type": "MultiPolygon",
        "coordinates": [
            [
                [
                    [-122.51, 37.69],
                    [-122.35, 37.69],
                    [-122.35, 37.80],
                    [-122.51, 37.80],
                    [-122.51, 37.69]
                ]
            ]
        ]
    },
    "subGeometry": {
        "type": "Polygon",
        "coordinates": [
            [
                [-122.49, 37.71],
                [-122.44, 37.71],
                [-122.44, 37.77],
                [-122.49, 37.77],
                [-122.49, 37.71]
            ]
        ]
    }
}

Example response:

See pipeline initiation.

Max zoom

Max zoom is determined by the following formula:

\[z_{max} = \lceil log_2(\frac{4 \cdot 10^7 \cdot cos(\frac{l \cdot \pi}{180})}{256 \cdot r_e}) \rceil\]

where \(l\) is latitude of the scene footprint centroid and \(r_e\) is the effective resolution in meters per pixel.

Effective resolution for the map can be determined by the following pseudo code:

if release_type == 'imagery':
    r_e = min(r for b in bands for r in (b.approx_res_x, b.approx_res_y))
elif provider == 'pl':
    r_e = 3.125
else:
    r_e = 0.5

Needed Permissions: kraken.release

POST /kraken/release/{map_type}/geojson/retrieve

Example request:

See pipeline retrieve.

Response JSON Object:
 
  • mapId (string) – map ID to be used to access map tiles.
  • maxZoom (int) – maximum zoom available for returned mapId.
  • tiles (list) – List of tiles which were successfully created, the system may be unable to release some of the requested tiles. This is always a subset of requested tiles. It is guaranteed that at most four tiles, all at equal zoom level, are returned. User is allowed to request any descendant tile, note that some child tiles of tiles may be missing.

Example response:

{
    "mapId": "xyz",
    "maxZoom": 19,
    "tiles": [
        [19, 83861, 202628]
    ]
}

Release Pairwise Map Types

General Change Detection and WRUNC Change map types use two scenes as input and are released and retrieved in the following manner.

Needed Permissions: See generic initiate endpoint.

POST /kraken/release/<map_type>/geojson/initiate
Request JSON Object:
 
  • oldSceneId (string) – required; ID of an old scene.
  • newSceneId (string) – required; ID of a scene in which changes will be detected as compared to “old scene”.
  • extent (object) – required; area of interest which must intersect with intersection of old and new scene’s footprints. The area must be allocated using Credits API if the imagery is high-resolution.
  • subGeometry (object) – optional; Subset of extent of interest. If provided the response contains subGeometryId which can later be used in Kraken Grid endpoint to crop result of analysis. See pipeline retrieve.
POST /kraken/release/<map_type>/geojson/retrieve

See generic retrieve endpoint.

Release with subGeometry

Sometimes it is useful to have kraken analyze bigger extent and then download only subset of what was analyzed. This is the case for SAR change algorithms when we want to analyze AoI buffered sufficiently so that kraken can use AoI neighborhood for data normalization.

Kraken Pairwise API

Use this endpoint to generate pairs from a list of scenes according to chosen pairing type and other options. Each pair can be afterwards analyzed as the pairwise map types.

Needed Permissions:

  • imagery.availability
POST /kraken/pairwise/initiate
Request JSON Object:
 
  • sceneIds (list) – required; IDs of scenes from which the pairs should be generated.
  • extent (object) – required; area of interest which must intersect with footprints of both scenes.
  • periodicity (string) – optional; Periodicity used to generate slots. Supported periodicities are “daily”, “weekly”, “amweekly”, “monthly” and “yearly”.
  • minIntersection (float) – optional; Minimum required intersection of the scenes in sq km. Defaults to 0.
  • pairingType (float) –

    optional; Possible options:

    • continuous create pairs from all successive records. Default value.
    • first pairs the first record (sorted by time of the occurrence) in the input to the rest of the records.
    • yoy pairs records with one year of time between them with some leeway to mitigate seasonality, see –max-yoy-delta.
  • maxYoyDelta (float) – optional; Maximum deviation from exact year over year pairing in days. Use decimal number for more accuracy. Only applicable for yoy pairing. Defaults to 15.0.
  • minDayDelta (float) – optional; SAR specific. Minimum span between paired scenes in days. Defaults to 0.0.
  • maxDayDelta (float) – optional; SAR specific. Maximum span between paired scenes in days. Defaults to unlimited day difference.
  • onlyTopOrbitNumber (bool) – optional; SAR specific. Whether to include only scene set with same relative orbit number and orbit pass that has most occurrences in input. Defaults to False.

Example request:

{
    "sceneIds": ["scene_id1","scene_id2","scene_id3"],
    "periodicity": "daily",
    "extent": {
        "type": "Polygon",
        "coordinates": [
            [
                [
                    [-122.51747131348, 37.694823535365],
                    [-122.35130310059, 37.694823535365],
                    [-122.35130310059, 37.809919574016],
                    [-122.51747131348, 37.809919574016],
                    [-122.51747131348, 37.694823535365]
                ]
            ]
        ]
    }
}

Example response:

Request JSON Object:
 
  • pairwise (list) – list of paired scenes. Dictionary contains oldSceneId and newSceneId.
{
    "pairwise": [
        {"oldSceneId": "scene_id1", "oldSceneId": "scene_id2"},
        {"oldSceneId": "scene_id2", "oldSceneId": "scene_id3"}
    ]
}

Grid Files

A tile can be downloaded using the following endpoint.

GET /kraken/grid/<map_id>/<geometry_id>/<z>/<x>/<y>/<file_name>
Parameters:
  • map_id (string) – ID of a map returned by the retrieve endpoint.
  • geometry_id (string) – For internal use only, use -.
  • z (int) – Zoom level of the tile.
  • x (int) – Horizontal coordinate of the tile.
  • y (int) – Vertical coordinate of the tile.
  • file_name (string) – One of the output files available for the released map type that should be returned. See the documentation of individual Kraken map types above.
Query Parameters:
 
  • force_download (int) – Set to 1 to download chosen output file even if the file has all pixels invalid.
  • download_unclipped (int) – Set to 1 to disable clipping of tiles to requested geometry in the Grid endpoint. kraken.download-unclipped permission needed.
Status Codes:
  • 200 OK – File is present and has valid data.
  • 204 No Content – PNG/SKI does not have valid pixels in requested area, this behaviour can be overridden by setting force_download to 1.
  • 400 Bad Request – Invalid request: a possible reason is invalid or expired map ID.
  • 401 Unauthorized – Given map ID doesn’t include requested tile.
  • 404 Not Found – Requested tile doesn’t exist: data are not available or the area of the tile cannot be analyzed.

Example requests:

GET /kraken/grid/xyz/-/19/83861/202628/truecolor.png HTTP/1.1
GET /kraken/grid/abc/-/19/83861/202628/aircraft.png HTTP/1.1
GET /kraken/grid/abc/-/19/83861/202628/segmentation.ski HTTP/1.1
GET /kraken/grid/abc/-/19/83861/202628/detections.geojson HTTP/1.1
GET /kraken/grid/abc/-/19/83861/202628/metadata.json HTTP/1.1

.png tile can be either RGBA or grayscale+alpha PNG and requested polygon is drawn to the PNG alpha channel.

truecolor.png can be represented by:

  • Bands corresponding to RGB in satellite imagery if such bands are present.
  • Bands chosen to fill RGB channels that are not corresponding to red, green, blue. For example imagery from idaho-swir contains only data from infrared bands and 3 of them are chosen to be visualized.
  • One Grayscale channel if only one band is present in satellite imagery. This grayscale band can correspond to visible light band in some imagery but generally it can contain other data too.
  • SAR imagery is visualized as (log(S), log(O), arctan(S/O)) where S is intensity of detected radio waves with same polarisation as radio waves emitted by the satellite, O is intensity of detected radio waves with polarisation perpendicular to emitted radio waves. arctan in blue band represents measure of preservation of polarisation plane in reflected light.

Kraken Dry-Run

Use this endpoint to get the size of ingested imagery and analysis. Useful for predicting costs of a particular analysis before it’s run.

Needed Permissions:

  • admin.tasking.get-pipeline
POST /kraken/dry-run/initiate
Request JSON Object:
 
  • extent (object) – required; area of interest which must intersect with scene’s footprint.
  • dryRuns (list) – required; list of dry run dictionaries. Contains list of scene dictionaries and a list of Kraken map types to use.

Example request:

{
  "dryRuns": [
    {
      "scenes": [
        {"sceneId": "DXc_fDD2aRjkHVOJ4elbKQ"},
        {"sceneId": "DXc_fDD2aRjkHVOJ4elbKg"}
      ],
      "mapTypes": ["imagery", "cars", "wrunc"]
    },
    {
      "scenes": [
        {"oldSceneId": "DXc_fDD2aRjkHVOJ4elbKQ",
         "newSceneId": "DXc_fDD2aRjkHVOJ4elbKg"}
      ],
      "mapTypes": ["wrunc-change", "change" ]
    }
  ],
  "extent": {
    "type": "MultiPolygon",
    "coordinates": [
      [
        [
          [-122.51747131348, 37.694823535365],
          [-122.35130310059, 37.694823535365],
          [-122.35130310059, 37.809919574016],
          [-122.51747131348, 37.809919574016],
          [-122.51747131348, 37.694823535365]
        ]
      ]
    ]
  }
}

Example response:

Request JSON Object:
 
  • ingested (float) – square kilometres of area that would be downloaded from the imagery provider.
  • analyzed (float) – square kilometres of area that would be analyzed.
  • kraken (float) – square kilometres of area that would be moved from Ragnar to Kraken.
{
    "ingested": 100.69,
    "kraken": 75.434,
    "analyzed": 50.23
}