Example Payload for Using the Coregister Tool

The coregister tool allows you to coregister a set of target items to a single anchor item within your order, making it easier to perform time series analysis of deep temporal imagery stacks. The coregistration tool ensures that images in a specified time series are spatially aligned, so that any feature in one image overlaps as precisely as possible with its position in any other image in the series.

This tool is designed to support coregistration of small areas of interest – contained within a single scene – and works best with high geographic overlap between scenes in the times series.

The source_type for this tool is scenes. The coregister tool supports all item types except for REScene and all bundle types except for basic_* and *_nitf bundles. It is not compatible with the composite too.

• anchor_item (string) (required): The item_id of the item to which all other items should be coregistered. Currently, only one item_id may be supplied, and it must be included as one of the products in the order. Items in the order must geographically overlap with the anchor item to be successfully coregistered.

{
  "name": "coreg_example",
  "products": [
    {
      "item_ids": [
        "20200720_194019_0f4c",
        "20200714_165944_0f4e",
        "20200630_164149_0e3a"
      ],
      "item_type": "PSScene",
      "product_bundle": "analytic_udm2"
    }
  ],
  "tools": [
    {
      "coregister": {
        "anchor_item": "20200630_164149_0e3a"
      }
    }
  ]
}

One imagery output file is produced for each bundle. The anchor imagery file and corresponding udm will have "_anchor" appended to their file names. Imagery files and their corresponding udms which are successfully coregistered, or already spatially aligned with the anchor item (i.e. did not need to be coregistered) will have "_coreg" appended to their file names.

The tool delivers an additional coregistration quality json file for each item in the order (except for the anchor item), which includes details on each output item's transformation. Rpc and xml files will be passed through.

Item coregistration quality JSON file parameters

The item coregistration quality json sidecar file includes the following information on coregistration transformations:

• projection_epsg_before: Lists the original projection of the item.
• projection_epsg_after: Lists the projection of the item after transformation. If it was transformed to the anchor item's reprojection, it will be updated. If the item was not reprojected, this field will be empty.
• pixel_shift: Describes the coregistration transformation in pixels.
• matching_score_before: A derived metric which describes image alignment before coregistration.
  • The matching score looks at a combination of Normalized Root Mean Square Error and Structural Similarity Index.
  • A matching score goes from 0 to 1.
  • A score above 0.8 indicates that the item already has very good alignment and is already correlated. These files will be marked with the "_coreg" suffix and will be passed through (or reprojected to the anchor item projection and delivered).
  • If the score is below 0.8, the tool will attempt to coregister the item. Note that a score of 0.8 does not necessarily indicate poor alignment.
• matching_score_after: A derived metric which describes image alignment after coregistration. If it was not coregistered because it was already correlated or it failed to successfully coregister, this field will be empty.
• matching_score_improvement: The difference in image alignment after coregistration.

Items which fail to coregister, or which result in a matching score improvement of less than 0.001 after the coregistration transformation will not be transformed. Instead, those will be delivered without any coregistration (or reprojection) transformation and "_coreg" will not be appended to their file names.