By Topic

Non-Parametric Object-Based Approaches to Carry Out ISA Classification From Archival Aerial Orthoimages

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Ismael Fernández Luque ; Department of Engineering, Area of Graphic Engineering, University of Almeria, Almería (Almería), Spain ; Fernando J. Aguilar ; M. Flor Álvarez ; Manuel Ángel Aguilar

In order to map the impervious surfaces for a coastal area, three non-parametric approaches: Classification and Regression Trees, Nearest Neighbor (NN), and Support Vector Machines (SVM)- were applied to a dataset of very high resolution archival orthoimages which had poor radiometry, with only red, green and blue spectral information. An object-based image analysis was carried out and four feature vectors were defined as input data for the classifier: 1) red, green and blue spectral information plus four relative spectral indices; 2) Dataset 1 plus texture indices based on the grey level co-occurrence matrix (GLCM); 3) Dataset 1 plus texture indices based on the local variance; and 4) the vector defined by 1), 2) and 3). Two classification strategies were developed in order to identify the pervious/impervious target classes (aggregation of all the subclasses and binary classification). The separability matrix was used to present the statistical comparative results clearly and concisely. The results obtained from this work showed that 1) “GLCM” texture indices did not lead to more accurate results; 2) the incorporation of the local variance texture index significantly increased the accuracy of the classification; 3) the classification results were not significantly affected by the classification strategy employed; 4) SVM and NN achieved statistically more accurate classification results than CARTs; 5) the SVM classifier was more efficient than the NN classifier, while NN was less dependent on the feature vector, and 6) suitable accuracy results were obtained for the most accurate approaches (SVM) which achieved a 89.4% overall accuracy.

Published in:

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing  (Volume:6 ,  Issue: 4 )