LEOSITS Work plan

Project’s goal is to accede beyond state of the art in SITS analysis by developing a set of methods and algorithms to describe the evolution of different categories (natural or manmade) in a scene.


LEOSITS is a three year project organized around 4 technical work packages as follows.

Work breakdown structure


WP0 – Project Management and Reporting - on going

 Main tasks include: Schedule management, Communication and reporting, Organization of reviews, Cost control and risk management, Documentation management.


WP1- Spatio-temporal primitive feature extraction - complete

The main objective was to structure SITS by expressing their content in a series of descriptors able to provide the best characterization of the areas inside. In addition, features extraction techniques were used for data dimensionality reduction.

To this purpose feature extraction methods have been studied and implemented considering first a single image (spectral indices, PCA analyis and Tasseld Cap transform) and then, pairs of images in order to asses correlation and Kullback Leibler divergence based similarity measures. 


WP2 - Image Time Series Change map generation - complete

Main tasks includes: Concept definition, development and implementation of change detection algorithms, Changes Map Time Series Generation and Optimization.

 Each pair of consecutive images from SITS was described by a number change maps computed using the different similarity measures. Complementary information about the changes occurred in the scene was extracted and employed in order to provide the user with a broader perspective on the land transformation processes. The resulting change maps are grey level images with the intensity of the pixel defining the degree of change.  


 WP3 Latency analysis of dynamic evolution-  on going

Main tasks include: SITS Model generation for dynamic evolution assessment. Individual category latent analysis


WP4 – Validation and evaluation -not started

 Previous generated models for dynamic evolution analysis of SITS will be compared with CORINE Land Cover classification. The main difference is that while CLC defines a number of four static classes, the proposed algorithm is able to identify a wide range of classes with dynamic evolution.