24, July 2014

PROJECT TITLE: Data Mining in Satellite Image Time Series using Constrained Sequential Pattern Extraction – COSEPE 

Coordinator: Institute of Space Science (ISS)  

Partners: – 

Period: November 2013 – November 2015

Project director: Andreea Maria Julea

Project team:

  • Project Manager (PM): Cristian Dumitru Ionescu (ISS) (former Andreea Julea, currently owing a postdoctoral research grant abroad)
  • Strategic Scientific  and Management Committee (SSMC): Catalin Cucu-Dumitrescu (ISS), Delia Teleaga (ASRC), Teodor Julea (ISS)
  • Executive Team (ET): Dan Selaru,  Veta Ghenescu (ISS) and Florin Serban,  Sorin Constantin (ASRC)

Description: Earth Observation (EO) – valuable tool for the land monitoring 

increase of data resolution (spatial, spectral and temporal), of dimensionality and complexity of data and attributes huge search space 

Data Mining process – automatically extracts interesting and potentially useful patterns or information from huge amount of data 

Satellite Image Time Series (SITS) – sequences of  pixel value evolution are criteria for characterization, discrimination and identification and thus, monitoring Land Cover and Land Use (LC&LU) entities

3D pixel: radiometric (value), spatial (location and connection) and temporal (order) 

The system provided for in the project 

uses Sequential Data Mining techniques (core of Knowledge Discovery in Databases)

is at pixel level and unsupervised;

the limitations of exponential search space (with specialization) and of too large solution set are exceeded  by dimensionality reduction and use of constraints

contains modules corresponding to the steps of a KDD process (preprocessing, sequential pattern extraction, post-processing for interpretation & evaluation)

Project objectives:

  • Development of a new system for LC&LU entities characterization based on an efficient constrained sequential pattern extraction from SITS implied by the use of anti-monotone connectivity measures. 
  • Development of new preprocessing and post processing techniques to improve the quality of extracted CoSP analysis and evaluation.


  • 1  Gathering user requirements and study on the status and trend of the SITS analysis
  • 2  Developing of LC&LU characterization system from SITS 

2.1. Developing of the preprocessing module

2.2. Developing of the pattern extraction module

2.3. Developing of the post-processing module

  • 3  Testing of operational capability of developed system and study of the influence of input data types and parameters variation

3.1. Testing of the operational capability of the developed system

3.2. Study of the influence of input data types and parameters variation on the extracted patterns

  • 4  System experimentation on real SITS and results evaluation using internal tests and feed back from potential users

4.1. Measure the performance of the procedures  

4.2. Corrections and calibrations of the system based on the feed back form users 

Contributions to the STAR programme objectives:

  • in line with the objectives stated within the Space Technology and Advanced Research (STAR)  Program  in the field of  Earth Observation
  • an important contribution to the Romanian Space and Security Research Program, especially in the research area of Space Applications, for the topics related to data acquisition and processing,  and automatic algorithms for spatial data          

Homepage: COSEPE