The main objectives of the project are:

• Development of a standalone prototype software tool to assess the geo-location accuracy of moderate to high resolution satellite sensors for
• The analysis of the geo-location stability of ESA optical sensors by analysing the geo-location accuracy of time series of EO data acquisitions
• Detection of possible geo-location errors
• Demonstration and evolution of the software tool for time series of ESA moderate resolution optical sensors

The project consists of the following Work Packages:

• WP-100: State of the Field
• WP-200: Definition of Algorithms, Processing lines and Identification of Existing Reference Data
• WP-300: Preparation of Reference Data
• WP-400: Development of Prototype Tool
  • WP-410: Prototype Tool V1
  • WP-420: Prototype Tool V2
• WP-500: Processing of 3 year data set and evaluation of geolocation accuracy
• WP-600: Project Management

Within a Contract Change Notice, the following sub-work-packages were added to the project:

• WP-310: Extension of reference data set to high latitudes
• WP-430: Prototype Tool V3 (support Proba-V)
• WP-510: Processing of remaining ENVISAT AATSR and MERIS FR
• WP-520: Processing of Proba-V Vegetation
• WP-700: Tools for analysis of Geolocation Shifts

The following Figure shows the processing line of the GeoAcca software:

1. Reference Data:
   A set of GCPs, consisting of lakes and islands, is used for the comparison between reference data and EO input data. Each GCP is defined by its edge coordinates, which are stored in a database. Cloud-free LANDSAT L1T scenes with a resolution of 30m are used as reference images and cut to the size of the GCP data windows.


2. Input Data:
   Level-1B EO input data (e.g. AATSR, MERIS) are orthorectified with the software BEAM and cut to GCP data windows slightly smaller than the reference data windows. Additionally, the input images are resampled to the same resolution as the reference images (30m). In order to identify and omit input data windows with cloud or snow coverage, a cloud and snow screening algorithm is implemented. If more than 10% of the pixels of an input data window are classified as cloud or snow covered, the image is excluded from further processing.


3. Template Matching:
   The near infrared (NIR) channel is used for the comparison of the reference image and the EO input image. The shift between the two images is calculated by template matching. Thereby the smaller input image is shifted pixel by pixel in comparison to the reference image: At each location, a correlation coefficient between the two images is calculated and stored in a matrix. The location with the best match can be retrieved from the maximum correlation coefficient.
   The following Figures show an example (GCP island Kea in Greece on 21.05.2004) of a LANDSAT reference image, an AATSR input image, the corresponding matrix of correlation coefficients and the AATSR image displaced by the calculated shifts. The yellow line shows the water outline from SRTM water body data.
   


4. Database of Geolocation Accuracy:
   The calculated shifts between input images and reference images are stored in the database together with metadata information on the input and reference images and geographic information on the respective GCP.

The following figures show timelines of the calculated geo-location shifts in along-track and across-track direction for all currently available sensors and GCPs:

For detailed analysis and evaluation of the calculated shifts between EO input images and reference images at the GCP locations, the online tool enables the creation of user-defined graphs based on the selection of different subdatasets and visualization modes. The online tool can be accessed in the Section Data Resources.

GeoAcca started in October 2013, with a duration of 18 months.
A Contract Change Notice started in March 2015, with a duration of 8 months.