Improving Land Cover Change Detection using Kernel Spectral Angle Mapper Approach in Hyperspectral Images
|
Mahdi Hasanlou * , Seyed Teimoor Seydi , Abdoreza Seydi |
University of Tehran |
|
Abstract: (4491 Views) |
Increasing the population and urban development is one of the most important human actions that cause changes on the face of the earth, especially in the developing countries, which is more. This process can cause devastating effects such as social, economic and biophysical. The harmful effects include; loss of agriculture lands, pasture and forest, change the pattern of the water, which somehow is associated with the changing patterns of land use and land cover. Land use and land cover changes as a basic factor in the changes of the Environment Act and converted into crisis. Identifying and evaluating the potential land-use patterns is essential, that if done on timely and with the high precision, it can help the planners and managers of relevant organizations for more conscious decision and making optimum use of resources in order to prevent the crisis. That would only be possible with the change detection. The hyperspectral images, due to having high spectral resolution, improved results of changes detection, provide more details of the changes. The main purpose of this research is to improve the process of land-use changes detection using spectral angle mapper algorithm, expectation maximization based on kernel based with hyperspectral imagery. The most important advantage of this method are as follow: unsupervised, no need to setting parameters of the knowledge basis, high precision and low false alarms rate. To evaluate the ability of the proposed method, hyperspectral imagery received from agricultural fields of Hermiston in the United States that captured by Hyperion sensors were used. The results are a significant improvement with the use of the proposed method for change detection in the standard spectral angle-mapping model compared to the top so that the overall accuracy is 94%, the coefficient Kappa 0.84 and false alarm rates of less than 6%. |
|
Keywords: Change Detection, Spectral Angle Mapper, Kernel Based, Hyperspectral Images, Expectation-Maximization Segmentation, Land Cover. Logistic Regression |
|
Full-Text [PDF 1205 kb]
(1266 Downloads)
|
Type of Study: Research |
Subject:
RS Received: 2016/11/29 | Accepted: 2017/11/12 | Published: 2018/06/21
|
|
|
|
|
Send email to the article author |
|