Title: Automated pool detection from satellite images using data mining techniques

Authors: Konstantinos Kontos; Manolis Maragoudakis

Addresses: Department of Information and Communication Systems Engineering, University of the Aegean, Samos, Greece ' Department of Information and Communication Systems Engineering, University of the Aegean, Samos, Greece

Abstract: A significant part of data mining is image mining, a principle that deals with image data relationship or patterns which are not explicitly found in the images. The basic objective is to gather knowledge from images with some known methods such as image retrieval, image processing and artificial intelligence. Towards this direction, this paper proposes a classification system for identifying pools from satellite area images. For the feature extraction of every photograph, and therefore the creation of the data set, the method of trainable segmentation is applied. In the current approach, genetic algorithms are utilised in an attempt to reduce the feature set to only the informative ones and class imbalance issues were also dealt with by incorporating a hybrid boosting and genetic sub-sampling approach. Results show that the best precision and recall rates are achieved by using a combination of AdaBoost and neural network algorithm.

Keywords: class imbalance; data mining; feature extraction; genetic algorithms; image classification; image mining; image processing; neural networks; object recognition; pool detection; satellite images; trainable segmentation.

DOI: 10.1504/IJIM.2016.083911

International Journal of Image Mining, 2016 Vol.2 No.2, pp.85 - 99

Received: 08 Oct 2015
Accepted: 17 Aug 2016
Published online: 26 Apr 2017 *

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