Multi Resolution Spatial Data Mining for Assessing Land use Patterns
Uttam Kumar1,  Chiranjit Mukhopadhyay2,   T. V. Ramachandra3,*

1Energy Research Group, Center for Ecological Sciences; Department of Management Studies; & Centre for Sustainable Technologies, Indian Institute of Science, Bangalore, India
2Department of Management Studies, Indian Institute of Science, Bangalore, India,

3Energy Research Group, Centre for Ecological Sciences; Centre for Sustainable Technologies; & Centre for Infrastructure, Sustainable Transport and Urban Planning (CiSTUP), Indian Institute of Science, Bangalore, India, Email:
Citation: Uttam Kumar, Mukhopahyay C and Ramachandra T V, 2015. Multi resolution spatial data mining for assessing land use patterns, Chapter 4, In Data mining and warehousing, Sudeep Elayidom (Eds), CENGAGE Learning, India Pvt Ltd., Pp 97-138.


Spatial Data Mining or Knowledge Discovery in Spatial Databases (KDSD), i.e. mining knowledge from large amounts of spatial data, acquired at various time intervals and from different sensors, is a highly demanding field. Spatial data mining helps in unraveling interesting and previously unknown, but potentially useful patterns from spatial datasets which is more challenging because of the diversity of multi sensor’s multi resolutions (spatial, spectral and temporal) data types with spatial relationships, autocorrelation, etc.

Spatial data acquired from multi resolution remote sensing sensors (in raster format) have been used to provide insights to landscape dynamics through comprehensive understanding of land use and land cover (LULC) pattern of a region. Detailed and accurate inventorying, mapping and monitoring of LULC at a local/regional scale have been possible with the availability of various medium to high spatial resolution sensors (such as Landsat, IRS LISS-III/IV, SPOT, IKONOS, etc.). The land use (LU) information are derived using image processing techniques based on the spectral properties of objects in the bands to assign them into a user defined class label.

Successful classification of spatial data into thematic information requires efficient and optimal image processing. Appropriate classification techniques and feature selection would enhance the classification accuracy. In this context, this chapter presents various advanced supervised pattern classification algorithms such as Maximum Likelihood Classifier, Decision Tree, K-Nearest Neighbour, Neural Network, Random Forest, Contextual Classification using sequential maximum a posteriori estimation, and Support Vector Machine, apart from their implementation on multi-resolution data acquired from different remote sensing satellites along with an assessment of the classifiers’ performance. Hybrid Bayesian Classifier a novel classification technique is presented with case studies in the last section of the chapter.

Keywords: Bayesian classifier, classification, machine learning, multi-resolution data, remote sensing, spatial data mining.


*Corresponding Author :
T.V Ramachandra,
Centre for Sustainable Technologies, Indian Institute of Science,
Bangalore 560 012, India.
Tel: 91-080-23600985 / 2293 3099/ 2293 2506, Fax: 91-080-23601428 /23600085 /2360685 (CES TVR).
E-mail   |   Sahyadri   |   ENVIS   |   GRASS   |   Energy   |   CES   |   CST   |   CiSTUP   |   IISc   |   E-mail