|
Application of genetic optimization algorithm in unsupervised classification and its comparison with ISO DATA and K-means classification methods
|
Ali Hossinholizade *1   , Najmeh Neysani Samany |
|
|
|
Abstract: (41 Views) |
| One of the most important methods of extracting information from remote sensing images is unsupervised classification, which allows users to generate a variety of information, including cover maps, land uses, and changes. This method does not require training data and the algorithm performs clustering operations according to statistical characteristics. In this research, in MATLAB coding environment with Landsat 8 images, ISO DATA, K-means and genetic algorithm methods were compared in parts of the north of Golestan province. For all three approaches, the conditions were considered completely constant to make a more appropriate comparison between the methods. In order to implement the genetic algorithm of 30 points (chromosomes), and to validate and check the accuracy of the results for each class, 10 points from Google earth images were used as a sample point. The results showed that the overall accuracy and kappa coefficient in unsupervised classification using genetic algorithm were 89 and 86%, K-means 54 and 47%, ISO DATA 48 and 42%, respectively. Therefore, it is suggested that considering the purpose and conditions of the research, the use of unsupervised classification using genetic algorithms, especially for areas of which sufficient information is not available, should be investigated. |
Article number: 6 |
| Keywords: Genetic Algorithm, Unsupervised Classification, Google Earth, Golestan Province |
|
|
Full-Text [PDF 1132 kb]
(24 Downloads)
|
Type of Study: Applicable |
Subject:
Special
Received: 2021/07/22 | Accepted: 2022/06/8 | Published: 2026/04/27
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
