A hyperspectral data cube is typically composed of about 100 to 200 spectral measurements for each spatial element of an imaged scene. They form the original set of the spectral features. From there, new linear or nonlinear transformed features can be generated in spectral space. In image space, spatial texture features, such as contrast, homogeneity and energy, can be derived and structure enhanced new features can be obtained by applying  Morphological filtering. 

Shape and size related features are available via object-based operations.Is creating more features better? Is using a large number of features in machine learning a good practice? In this talk, issues in effective features generation and selection will be addressed. An information class separability measure in cluster space will be introduced. The reduced subset of features is expected to minimize redundancies, enhance class separability and avoid the Hughes phenomenon. Feature integration approaches will be discussed briefly as well.

Xiuping Jia received the B. Eng. degree from Beijing University of Posts and Telecommunications, Beijing, China, in 1982 and the Ph.D degree in Electrical Engineering from The University of New South Wales, Australia, in 1996. Her thesis was titled ‘Classification techniques for hyperspectral remote sensing image data’. Since then she has continued her research in image processing, data analysis and remote sensing applications. The projects she has been involved in and supervised range from image registration, data compression, feature reduction, to spectral-spatial based classification. Subpixel mapping has been addressed in recent years via spectral unmixing techniques and super resolution reconstruction approaches. 

Dr. Jia is currently with the School of Engineering and Information Technology, University College, University of New South Wales, Australian Defence Force Academy, Australia. She was a visiting scholar at several organizations during her sabbatical leaves, including Purdue University, USA, Beijing Normal University, The Chinese University of Hong Kong, and The Chinese Academy of Sciences, China.

She is the co-author of the remote sensing textbook, Remote Sensing Digital Image Analysis, Berlin, Germany: Springer-Verlag, 3rd (1999) and 4th (2006) eds. She is as an Associate Editor of IEEE Geoscience and Remote Sensing, and a member of the International Committee for Imaging Science. Dr. Jia is a Guest Professor of Beijing Normal University.