Earth observation is currently developing more rapidly than ever before. During most of the last three decades satellite-based remote sensing has typically been accomplished by a few satellites with a single sensor on each and low temporal coverage. The last few years and the near future show a b ig difference. Satellite platforms with a large number of sensors are coming , and a huge number of satellites with one or a few sensors emerge. The coverage of the Earth in space, time and the electromagnetic spectrum is increasing correspondingly fast.

 

This development opens for a potential significant change in the approach of analysis of earth observation data. Traditionally, analysis of such data has been by means of analysis of a single satellite image. The emerging exceptional good coverage in space, time and the spectrum opens for analysis of time series of data, combining different sensor types, combining imagery of different scales and better integration with ancillary data and models. Classical methodology, like a Maximum Likelihood or Maximum Aposteriori classifier, will in general not be able to extract all the interesting information present in this type of data sets. New methods are needed which are optimised for that.

 

The project will develop a common mathematical framework for multi-data analysis and apply this for the development of methods for analysis of earth observation data from different sensors, with a high number of bands, with different spatial resolution and acquired over a period of time under changing conditions. The methods will be general, enabling use within a broad range of earth observation applications. The project will make these analysis tools available to the other projects in the Research Programme and actively seek collaboration in order to test the methods on various monitoring problems. The project will accomplish the work through the following work packages:

 

Establishing a common mathematical framework for the development of tools for multi-sensor, hyperspectral, multi-scale and multi-temporal analysis of satellite image data sets covering the same geographical area.

Develop a method for simultaneous analysis of a set of data from different sensors, including ancillary data.

Develop a method for hyperspectral data analysis selecting an optimal subset of bands and utilizing the variance in high-dimensional feature space.

Develop a method that combines data of various spatial resolutions and is able to utilize high-resolution information for calibration of low-resolution data analysis.

To develop a method for analysis of time series of data which is able to utilize models for spectral class development (e.g., a phenological vegetation model).

Verification and demonstration of methods with associated partners and other interested projects within the Research Programme.