Title: An approach to one-class extraction from remote sensing imagery

Authors: Shukui Bo; Yongju Jing

Addresses: Department of Computer Science and Application, Zhengzhou Institute of Aeronautical Industry Management, No. 2 of Daxue Middle Road, Zhengzhou City, Henan Province, 450015, China ' Library, Zhengzhou Institute of Aeronautical Industry Management, Zhengzhou 450015, China

Abstract: One-class extraction tries to distinguish a specific class of interest from the remainder of the data set. The one-class extraction problem can be solved by either a one-class classifier or a multi-class classifier. With a multi-class classifier, the one-class extraction of remote sensing image is studied in this article. First, we analyse the probability of error for one-class extraction with nearest neighbour classifier. As a non-parametric method, the nearest neighbour classifier requires the data distribution to be partitioned into only two classes, the class of interest and the remainder. Second, with the two-class partitioning of the dataset, the specific class of interest is well extracted from a remotely sensed image. This study improves the classification process and would be helpful for one-class extraction of remote sensing imagery with multi-class classifiers.

Keywords: nearest neighbour; one-class; remote sensing.

DOI: 10.1504/IJICT.2017.085465

International Journal of Information and Communication Technology, 2017 Vol.11 No.1, pp.119 - 127

Received: 29 Oct 2014
Accepted: 15 Nov 2014

Published online: 21 Jul 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article