FAST4MAP (Fast & Advanced SAR Techniques for Monitoring & Alerting Processes) is a project co-funded by the Italian Space Agency (ASI), whose goal is to improve the speed and accuracy of E.O. data processing, leveraging the GPU's processing capabilities. One of the first demonstration application is the creation of a dedicated landslide service, based on SAR data (Sentinel-1A/B, COSMO-SkyMed and, for the near future COSMO-SkyMed 2nd Generation).
Landslide monitoring service
The service uses mainly Sentinel-1A/B products for continuous monitoring of areas at high-risk of landslides, using COSMO-SkyMed (CSK) data at higher resolution for event based analysis (danger situation or alert prediction) by accessing the modality of planning and acquisition in emergency situations as provided for CSK mission. Moreover it provides an integration of Sentinel-1 and CSK results.
This service is based on capability to measure ground displacements from SAR interferometric data. In particular, the Multi-Temporal InSAR (MTI) techniques allow to accurately measure millimetric soil displacements in correspondence of those targets on the ground (PS - Persistent Scatters and DS - Distributed Scatters) that have a high temporal radar coherence. The user can subscribe a periodic service over a geographic area to be monitored and access through authentication on a web-GIS interface where is it possible to analyze the results of the MTI processing, speed monitoring and amplitude trends of PS and DS rendered in geo-analytic form.
The processing chain provides integration between PS and DS techniques (integration of Spinua-PS and DARIS-DS techniques) and exploits mainly Interferometric Wide Swath open and free data provided by Sentinel-1 mission, ensuring continuous monitoring, and Stripmap, Spotlight data provided by Cosmo-SkyMed. The MTI techniques are based on complex algorithms, whose configuration and execution are typically offloaded to expert operators, who can recognize and also manage abnormal situations due to particular noise conditions or in the presence of artefacts.
In case of FAST4MAP, the SPINUA-DARIS processing chain is fully automatic. Moreover, to speed up the processing, strongly heavy from the computational point of view, some processing algorithms exploit massive parallelism of GPU; therefore some code running on standard HW architecture CPU based, is re-written by kernels running on GPUs and considering a hybrid architecture host-device. This allows a compact architecture that can be easily moved in proximity of the data and with a best speed ratio than more traditional computing server. Fast4Map produced an open and free library under BSD license (libGPU4EO) for massive processing of EO data using GPGPU architecture, based on the OpenCL framework.
Contacts
Companies or scientific communities wishing to request the libGPU4EO library can contact the project manager at the email address agrimano @ planetek.it, giving the purpose of the request. In the framework of the Italian Space Agency's (ASI) "Quarto bando tematico: Navigazione e osservazione della Terra". Prime contractor: Planetek Italia srl; Partner: GAP srl.
