/

Land Lines Image Segmentation

The Land Lines Image Segmentation (LLIS) process block generates GeoJSON polygons that delineate major land cover classes, vegetation management zones, and natural land surface features from sentinel-2 satellite imagery.


Introduction

This block generates GeoJSON polygons that delineate major land cover classes and vegetation management zones from Sentinel-2 satellite images.

Technical Information

Algorithm description

The processing algorithm generates GeoJSON polygons that extract major land cover classes from Sentinel-2 satellite images, including lakes, rivers, vegetation cover and types (forest, shrubs, grassland, cropland, etc), and other visually distinct classes. The average polygon size is 10 ha for large segment outputs and 1 ha for small segment outputs. Features smaller than 10 pixels will not be delineated. The maximum area per run is equal to a Sentinel-2 full scene (~10,000 km2).

This algorithm was developed over several years of research by automating image segmentation workflows for the high accuracy identification of forest stands and optimal delineation using LiDAR and aerial imagery, in order to create uniform stand boundaries for precision forestry management (for more information, please refer to Tesera website).

This algorithm was trained on cloud-free Sentinel-2 images of natural landscapes that cover areas ranging from 10 km2 to 10 000 km2 all over the world including Canada, USA, Brazil, Chile, Europe, China, and India.

Output files

OutputFile format
Large landlines polygonsGeoJSON
Small landlines polygonsGeoJSON
Sentinel-2 RGB imageGeoTIFF

Compatible blocks

Compatible blocks (before)Compatible blocks (after)
Sentinel-2 L2A Analytic (GeoTIFF)Count Objects
Export Data (Vector)

Geographic coverage

The geographic coverage is global.

How it works

This block has no configurable JSON parameters.

Disclaimer

Known limitations: clouds, shadows, and snow will affect the quality of the results. Since the filter max_cloud_cover for the input imagery applies to the Sentinel-2 full scene and not just the user-specified image segment, we recommend setting this filter to 0% or less than 5% for best results.

The algorithm was trained primarily on Sentinel-2 imagery covering Canada, USA, Brazil, Chile, Europe, China, and India. The performance of this algorithm might be lower for images covering other regions.

Examples

Example of a workflow created with the data block Sentinel-2 L2A Analytic (GeoTIFF) and Tesera Land Lines Image Segmentation:

Loading...

Capabilities

Input

Output

To know more please check the block capabilities specifications.