Fully automated system extracts information on smallholder farms from raw satellite data
ITC in the last two years has strongly invested in the creation of an automated satellite image processing node for the needs of STARS. After several cycles of designing, implementing, debugging and establishing a collection of algorithms we are now confident that a fully-automated Very High Spatial Resolution (VHSR) satellite image processing system addressing the specifications and needs of smallholder farming remote sensing has been in place and can extract information for the farm management units under scrutiny in near-real time.
Fig 1: STARS has developed a fully-automated satellite image processing system that can extract relevant information for farm management units from Very High Spatial Resolution (VHSR) satellite data.
Practically, this is a development of a processing flow from raw satellite images delivered from Digital Globe to crop-related information assigned per individual smallholder farm plot. This development has been the outcome of the realization of the need to develop automated processes from the early stages of the project. The processing node is based fully on free software and largely on open source software, such as R, Python, the Orfeo toolbox, bash scripting etc. This workflow has been tested and applied on a multi-sensor image archive of over 270 VHSR deliveries of WorldView-2, WorldView-3, QuickBird and GeoEye data in five different test areas in South-East Asia and sub-Saharan Africa and is now customized for RapidEye data as well.
The workflow delivers two types of products (see figure 2). First, a full pre-processing chain produces image products typically needed in remote sensing studies; such product levels are the atmospheric correction, mosaicked images, ortho-rectified products, image-to-image geo-registered products, tree mask and cloud mask. Second, information specific to a farm management unit or a crop type during a crop season is generated such as the mean and variance of the reflectance of a field unit, several vegetation indices and textural characteristics which are ready in a data format or visualization figures, as for instance the temporal evolution of the NDVI over a specific crop in a growing season. All this statistically-extracted information is stored in a database, the so called spectro-temporal spatial database (CSSL). Moreover, the products are readily available in real-time from the STARS-dedicated server in ITC to the STARS partners for further manipulation and interpretation, and the CSSL information will be available to the general public after the project.
Fig 2: The STARS satellite image workflow delivers two types of products. First, a full pre-processing chain produces image products typically needed in remote sensing studies (white boxes on the right); Second, statistically-extracted information is stored in a database, the so called spectro-temporal spatial database (CSSL).
With an automated workflow coming available to the general public as a project outcome, producing these two types of products we hope to aid the timely pre-processing of the satellite imagery and:
1. make it easily available to the STARS partners who can further process a neat dataset
2. deliver farm- and crop-specific readable information to the public and agencies to aid their decision making and
3. make available the satellite image automated workflow through our Knowledge Portal in the course of this year and therefore support developing countries scientists who can base their agriculture remote sensing efforts on a free and open source collection of algorithms.