Tree detection

An algorithm that detects trees in SPOT, Pléiades, or Pléiades Neo imagery and returns a probability map.


Overview

The algorithm tiles the input imagery and identifies all pixels that are part of a tree in SPOT, Pléiades, or Pléiades Neo display imagery. The result is a GeoTIFF file that maps the probability of each pixel being part of a tree.

Tree detection can be used for infrastructure vegetation risk monitoring, urban planning, and construction.

An input SPOT image of Paihuano (Chile)

A tree probability map


See more on the marketplace.

Performance

The algorithm has been developed using tens of thousands of semantically annotated images.

The algorithm has achieved an F1 score of 93%. For optimal results, the algorithm must be used to detect medium or large trees in full foliage. You can use it to detect medium or large trees with less foliage, but the results might not be as accurate.

Requirements for input imagery

The data item must be a supported data product:

The data item must be CNAM-compatible. Check that the data item has been added to storage in 2023 or later.

Input parameters

You must specify the data items you want to apply the process to.

You must specify the title of the output data item.

API input

Use the detection-trees-spacept name ID for the processing API.

JSON
{
"inputs": {
"title": "Processing imagery over Berlin",
"item": "https://api.up42.com/v2/assets/stac/collections/21c0b14e-3434-4675-98d1-f225507ded99/items/23e4567-e89b-12d3-a456-426614174000"
}
}
ParameterOverview
inputs.title

object | required

The title of the output data item.

inputs.item

object | required

A link to the data item in the following format: https://api.up42.com/v2/assets/stac/collections/{collection-id}/items/{item-id}.


Last updated: