Verde Image processing description

The Overland processor is the core element of Verde

The Verde service uses the Overland processor, an optical image processing suite developed by Airbus to generate vegetation maps such as Leaf Area Index (LAI) or Leaf Chlorophyll content (CHL). Initial algorithms were developed in the early 2000's and have been constantly improved since that date. Overland is able to process a large range of multispectral images, covering spectral domains 0.4 to 2.5µ from various sources (satellite, airborne, UAV) and spatial resolutions.

Overland, how does it work?

The Overland processor uses techniques based on physical models, so called "biophysical processing". The SAIL and PROSPECT models are the core elements of the model simulating the reflectance of the crop canopy (Verhoef, 1984, Verhoef, 1985, Jacquemoud & Baret, 1990). The association of SAIL/PROSPECT models is highly successful and has been validated on many crops (Jacquemoud & al, 2009). Satellite observations passing through the whole column and the modelling of the atmospheric transfer is performed thanks to the LOWTRAN model (Kneisys et al., 1995) completed with a dedicated cloud model (turbid medium model using cloud optical properties and a Henyey-Greenstein phase function). In Verde, each part of these models is continuously being improved, in order to increase the overall robustness and accuracy. The scope is also regularly enlarged with new crops and new sensors added to the processing heart. The user can find the version of the modeling that has been applied to his / her subscriptions in the metadata of his / her packages (e.g. “FM02”).

The Overland processing principle is to couple the scene and atmospheric models in order to perform inversion of this combined model through minimization techniques, having the satellite image converted to Top of Atmosphere (TOA) radiance as inputs. Advantages of such an approach are discussed in (Verhoef & Bach, 2003). A detailed description of the Overland algorithms can be found in the Algorithm Theoretical Basis Document (ATBD) of the geoland2 MERIS products (Poilvé, 2010). In this case, they were applied to process low-resolution MERIS data (15 VNIR bands / 300 m).

Overland and Verde

The Verde service uses publicly available imagery that best fits the Agriculture application, i.e. Sentinel 2A/2B and Landsat 8, as well as Airbus images (SPOT 6/7 and Pléiades 1a/1b) in order to improve spatial resolution and revisit. LAI maps can be produced from all these sources, whereas Chlorophyll maps are only provided from the Sentinel 2A/2B data. The spectral richness of the Sentinel 2A/2B sensors (13 bands) allow for robust retrieval of this information. The resolution of the processing is optimized depending on the user choice of resolution (‘Common’ or ‘Resolution’, see part II section Layers, sensors, cloud cover, resolution). This is achieved by a multi-resolution processing technique that performs spatial enhancement similar to a pan-sharpening technique, but integrated within the biophysical processing.

The Overland processor, with its built-in atmospheric model, performs autonomous atmospheric corrections and automatic masking of thick clouds and dark shadows. A map is discarded from the series of observations if field plot masked area exceeds a maximum fraction (e.g. 30%, customizable). Applied quality rules also lead to discarded conditions such as snow cover or a flooded field (except for rice).

As Verde uses imagery from various satellite sources, featuring different geometric performances and different native resolutions, a co-registration step has been added to the processing in order to ensure a perfect overlay of all observations within a time-series, regardless of their source sensor. It removes the shift one can observe when browsing a satellite image time series. This feature leverages a proprietary database of Ground Control Points (One Atlas Basemap) that Airbus has built globally and updates every year.

Verde processing chain has been migrated to the Cloud (Google Cloud Platform) and can be triggered through API calls. The infrastructure behind the service has been designed for maximum scalability and optimal performance, as the number of virtual machines activated is proportional to the amount of requests at a given point in time1.

Refences

  1. Jacquemoud, S. & Baret, F. (1990). PROSPECT: A model of leaf optical properties spectra. Remote Sensing of Environment, 34, 75-91.
  2. Jacquemoud, S. & al. (2009). PROSPECT + SAIL models: A review of use for vegetation characterization. Remote Sensing of Environment, 113, S56-S66.
  3. Kneisys F.X., Abreu L.W., Anderson G.P., Chetwynd J.H, & al., (1995). The MODTRAN 2/3 and LOWTRAN 7 Model. Philips Laboratory, prepared by Ontar Corporation, North Andover (MA), 267 pp.
  4. Poilvé, H. (2010). BioPar Methods Compendium of MERIS FR Biophysical Products (report g2-BP-RP-038, EC geoland2 project FP-7-218795). Retrieved from ResearchGate website: https://www.researchgate.net/publication/265728093_geoland2_-_BioPar_Methods_Compendium_of_MERIS_FR_Biophysical_ProductsBioPar
  5. Verhoef, W. (1984). Light scattering by leaf layers with application to canopy reflectance modeling: the SAIL model. Remote Sensing of Environment, 16, 125-141.
  6. Verhoef, W. (1985). Earth observation modelling based on layer scattering matrices. Remote Sensing of Environment, 17, 165-178.
  7. Verhoef, W., & Bach, H. (2003). Simulation of hyperspectral and directional radiance images using coupled biophysical and atmospheric radiative transfer models.  Remote Sensing of Environment, 87, 23–41.

Visit our biophysical parameters page to get the comprehensive list of research papers about this approach and its benefits:

 

Did You Find Your Solution?

Airbus is there to support your business. We are ready to deliver.