FD-7: Advanced Classification Techniques for Remote Sensing

Presented by: Ranga Raju Vatsavai and Surya Durbha
1 Oak Ridge National Laboratory, USA
2 Indian Institute of Technology, India

Supervised learning (classification) is the most widely used technique for thematic classification of remote sensing images [15]. Statistical pattern recognition algorithms, especially the maximum likelihood classifier is most extensively studied and utilized for classification of multi-spectral images. Decision trees and neural networks have also been widely applied for multisource (remote sensing images and ancillary geospatial databases) classification. However, recent advances in remote sensing technology, especially improved spectral, spatial, and temporal resolutions, puts several constraints on the traditional classification algorithms. For example, increasing spectral and temporal resolution requires large number of training samples (typically 10 to 30 times the number of bands), and increasing spatial resolution invalidates the fundamental assumption that the training samples are independent and identically distributed (iid). These advancements in remote sensing data have prompted the development of advanced classification algorithms such as semi-supervised learning, active learning, and spatial classification to overcome some of the challenges posed by new datasets. Unfortunately these new algorithms have been confined mostly to the academic researchers, and the current commercial implementations are oblivious to these advances. This tutorial tries to fill this gap by bringing the recent advances to the practitioners.


The primary objective of this tutorial is to bring recent advances in classification technology to the remote sensing analyst. Through this tutorial we would like to disseminate the basic principles behind these new classification and machine learning schemes, and give the participants a firsthand practical experience through open source advanced classification tools.


(1) Statistical Framework: We first introduce basic concepts, such as, maximum likelihood parameter estimation, Bayesian classification framework, Gaussian Mixture Models, Expecta- tion Maximization, Covariance structure and robust estimation techniques.

(2) Semi-supervised learning techniques that utilize large unlabeled training samples in conjunction with small labeled training data are becoming popular in machine learning and data mining. This popularity can be attributed to the fact that several of these studies have reported improved classification and prediction accuracies, and that the unlabeled training samples comes almost for free. We will introduce basic semi-supervised learning framework for the classification of remote sensing imagery.

(3) Sub-class Classification: Increased spectral resolution offers the remote sensing analyst the ability to carryout species level classification, however it also requires additional training efforts. We will introduce a new sub-class classification scheme that is capable of automatically identifying finer (sub-) classes from aggregate classes, thus reduces the need for large amounts of additional training data.

(4) Large-Margin Classifiers: We discuss the concepts underlying large margin classification and introduce popular support vector machines classification algorithm.

(5) Spatial Classification: Increasing spatial resolution invalidates the basic assumption that the training samples are independent and identically distributed. We introduce a spatial semi-supervised learning scheme and also allude the participants to the basic differences with respect to the conventional per-pixel based classification schemes. We discuss conventional Markov Random Fields and as well as recent Gaussian Process (GP) based classification.

(6) Multiple Classifier Systems: We introduce basic premise behind multiple classifier systems and discuss various classification fusion schemes. Students will learn how to combine outputs from the classification schemes learned above and evaluate the results.


Ranga Raju Vatsavai is a senior research scientist at the Oak Ridge National Laboratory. He has published over 80 peer-reviewed articles, co-edited two books on "Knowledge Discovery from Sensor Data" and a special issue of Intelligent Data Analysis Journal (Volume 13(3), 2009), co-authored "Geographic Data Mining and Knowledge Discovery" research priority for UCGIS, served on program committees of several international conferences including ACM-KDD, SDM and ACMGIS, conducted tutorials on advanced classification techniques for remote sensing (Pecora 2008, IGARSS: 2009, 2010, 2011) and co-chaired SensorKDD (with ACM-SIGKDD: 2007-2011), SSTDM (with IEEE ICDM: 2008-2011), KDCloud (with IEEE ICDM: 2010, 2011), PDAC (with ACM/IEEE Supercomputing: 2010, 2011), and HPDGIS (with ACM SIGSPATIAL GIS: 2010, 2011), and LDMTA (with ACM-SIGKDD 2011) workshops, co-organized invited sessions on data mining and machine learning for remote sensing at the IEEE IGARSS (2009, 2010, 2011). His research interests include Spatial and Spatiotemporal data mining, machine learning, remote sensing, and GIS. He received his MS and PhD in computer science from the University of Minnesota, USA.

Surya S. Durbha is an Assistant Professor at the Indian Institute of Technology. Previously he worked as Assistant Research Professor at the Center for Advanced Vehicular Systems (CAVS) and also held an adjunct faculty position with the electrical and computer engineering department at Mississippi State University. He received his M.S. degree in remote sensing from Andhra University, India, in 1997 and Ph.D degree in computer engineering from Mississippi State University (MSU), MS, U.S.A, in 2006. Earlier he was an application scientist at the Indian Institute of Remote Sensing, Department of Space, India (Dehradun, 1998 to 2001) and also worked in Rolta India limited (Mumbai, 1997). He is currently working in the area of image information mining tools for content-based knowledge retrieval from remote sensing imagery and in the recent past worked on the retrieval of biophysical variables from multi-angle satellite data. He has published over 30 peer reviewed articles, served on program committees of several international conferences including SSKI, SSTDM, and IGARSS, and co-chaired sessions at various conferences. His current research interests are semantics, knowledge-based systems, image information mining, remote sensing, and sensor webs.