Methodology of Multitemporal Multisensor Satellite Image Analysis

This project objective was to systematically evaluate existing satellite image time series (pre)processing approaches and develop a general processing time series analysis workflow for various applications, combined with the in-depth study of the impact of time series data properties (like radiometric features and density) on time series analysis and applications.

The final reliability and usefulness of time series results for a selected application depend on the data quality, the length and density of time series and the selected processing methods. In the project, we focused on the multi-sensor harmonization, the generation of time series and validation of the most critical processing steps and their effect on final quality of results. The consistency of the data is one of the main prerequisites in time series analysis, therefore calibration of sensor and image differences and harmonization of time series are very important tasks.

In terms of this perspective we had five objectives:

• Development of a generic method for multi-sensor data calibration,
• Harmonization of different sensor data into a single, long and dense time series,
• Application of (massive) cloud data processing,
• Systematic evaluation of the latest time series analysis algorithms (software tools), and
• Observation of selected events and phenomena in Slovenia.

Numerous experiments were performed to reach those objectives. However, after thorough research we can present the most important results that include:

• development of general methods for data reconciliation between sensors,
• harmonisation of data from different sensors into a single, long and dense time series,
• transition to cloud-based processing,
• systematic evaluation of state-of-the-art algorithms for time series analysis and
• observation of selected events and phenomena in Slovenia.

More information about the project can be found at the leading partner web page.