IMPLEMENTATION CASE-STUDY OF AUTOMATED CHAINS FOR LEVEL-3 SATELLITE IMAGERY GENERATION
The paper describes an automated image processing chain implementation case-study. The objective of this chain is to obtain level-3 satellite imagery from raw observations with minimum human intervention. The chain also generates several high level products. The chain is controlled via web interface which includes visualization of the intermediate steps.
Meris raw imagery is downloaded daily from provider ftp servers. The data is selected using a geographical bounding box. First a segmentation stage using object-oriented fuzzy classifier is applied to extract clouds and water bodies. Then an automated registration process using common point features extracted form non cloud or water areas is used to adjust a mathematical model. Using the registration information the image is reprojected to a common reference frame. Optionally atmospheric correction algorithms can be applied. Finally the user can generate and visualize mosaics, vegetation indexes and principal component analysis using a web 2.0 based interface.
OSSIM libraries provide support for raster analysis. The process is parallelized across a beowulf class cluster using OSSIM MPI support. Custom plugins warping OpenCV functions have been developed and released with GPL license. Process automation is done using bash scripts triggered via php5 and crontab. The web server uses cgi-mapserver as wms provider and the interface is developed with Openlayers.
Jorge Artieda - Argongra
Manuel Aymerich - Argongra
Rebeca Gutiérrez - Argongra