You are at INPE :: Earth System Science Centre :: Geospatial Change Modelling Research :: Geospatial Change Modelling Techniques :: Geographical Data Mining

Geographical Data Mining

Remote sensing provides information about land use over large areas. and information extraction in remote sensing databases requires adequate methods. The aim of this research is to develop methodologies and computer tools for extraction of information in remote sensing image databases. We focus on the automatic identification of different types of land change objects and on reconstructing an object´s life from snapshots provided by a series of remote sensing. images.

The above figure shows an analysis of the evolution of land change objects of different types in Vale do Anari, Rondonia. The results demonstrate that a clear trend towards land concentration.

Research Team

The Geographical Data Mining research at INPE is led by Dr.Gilberto Câmara, Dra. Leila Fonseca and Dr. Isabel Escada, and PhD students Joice Seleme Mota, Olga Oliveira, Thales Korting. We cooperate with groups at the Universidade Estadual do Rio Grande do Norte (Marcelino Pereira).

Research Results

We have proposed a methodology for pattern classification of land change objects based on landscape ecology metrics Remote Sensing Image Mining: Detecting Agents of Land Use Change in Tropical Forest Areas. M.P.S. Silva, G. Câmara, M.I. Escada, R.C.M. Souza. International Journal of Remote Sensing,vol 29(16): 4803 – 4822, 2008.

We have also proposed a conceptual view of an object´s evolution. We consider cases where the evolution of an object is dependent of its type and propose a rule-based approach for description of spatiotemporal object evolution. See Rule-based Evolution of Typed Spatio-temporal Objects. O.Bittencourt, G.Câmara, L.Vinhas, J.Mota. IX Brazilian Symposium in Geoinformatics, GeoInfo 2007, Campos do Jordão, 2007.

We have also shown how a case-based reasoning technique can be used to establish the evolution rules for spatiotemporal objects. See "Applying Case-Based Reasoning in the Evolution of Deforestation Patterns in the Brazilian Amazonia”, by Joice Mota, Gilberto Câmara, Leila Fonseca, Maria Isabel S. Escada, and Olga Bittencourt. 23rd Annual ACM Symposium on Applied Computing. Fortaleza, Brasil 2008.

Computer Tools

We also developed a tool called Geographical Data Mining Analyst – GeoDMA, which runs as an add-on for TerraView . In its first version, GeoDMA is a pattern classifier for land objects. The input data is either a set of binary objects with different shapes and/or a set of images. The user then selects the set of shape metrics to be used for classification. The training phase uses the C4.5 data mining algorithm to establish a decision tree, which is then applied to the input data.

For more details, see the GeoDMA Page.

QR Code
QR Code geomining (generated for current page)