A Spectral Unmixing Model for the Integration of Multi-Sensor Imagery: A Tool to Generate Consistent Time Series Data

Author： Doxani Georgia Mitraka Zina Gascon Ferran Goryl Philippe Bojkov Bojan R.

Publisher： MDPI

E-ISSN： 2072-4292|7|10|14000-14018

ISSN： 2072-4292

Source： Remote Sensing, Vol.7, Iss.10, 2015-10, pp. : 14000-14018

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

The Sentinel missions have been designed to support the operational services of the Copernicus program, ensuring long-term availability of data for a wide range of spectral, spatial and temporal resolutions. In particular, Sentinel-2 (S-2) data with improved high spatial resolution and higher revisit frequency (five days with the pair of satellites in operation) will play a fundamental role in recording land cover types and monitoring land cover changes at regular intervals. Nevertheless, cloud coverage usually hinders the time series availability and consequently the continuous land surface monitoring. In an attempt to alleviate this limitation, the synergistic use of instruments with different features is investigated, aiming at the future synergy of the S-2 MultiSpectral Instrument (MSI) and Sentinel-3 (S-3) Ocean and Land Colour Instrument (OLCI). To that end, an unmixing model is proposed with the intention of integrating the benefits of the two Sentinel missions, when both in orbit, in one composite image. The main goal is to fill the data gaps in the S-2 record, based on the more frequent information of the S-3 time series. The proposed fusion model has been applied on MODIS (MOD09GA L2G) and SPOT4 (Take 5) data and the experimental results have demonstrated that the approach has high potential. However, the different acquisition characteristics of the sensors, i.e. illumination and viewing geometry, should be taken into consideration and bidirectional effects correction has to be performed in order to reduce noise in the reflectance time series.