Application Of Support Vector Machines To Analyze Ambient Particulate Matter (pp. 75-102)
Authors: (K. Manivannan, V. Devabhaktuni, A. Kumar, P. Aggarwal, P. Bhattacharya, EECS Department, University of Toledo, Toledo, Ohio, USA, and others)
Abstract: This chapter presents an efficient automated technique for the selection of optimal
segmentation algorithm to characterize the particulate matter (PM).
The PM images are captured using the scanning electron microscope (SEM)
embedded with energy dispersive x-ray (EDX) spectroscopy.
Support vector machines (SVMs) is the classification tool employed for its reliable
and accurate prediction of optimal segmentation algorithm for each image. The feature
extraction and representation of the image for effective training of SVM is performed by
second order statistical method known as gray level co-occurrence matrix (GLCM).
The matrix is calculated at various angles and the texture features are evaluated to
classify the images.
The GLCM based SVM training illustrated drastically better performance than the
standard techniques due to the additional knowledge based on spatial relationship