|
Improving the urban features classification accuracy by fusion of optical and radar high spatial resolution images
|
Fattane Kia *   , Mohammad javad Valadan Zoej , Fahimeh Yousefi |
| K. N. Toosi University of Technology |
|
|
Abstract: (1306 Views) |
| The population growth and the development of the urban environments have created a lot of incentives for researchers in the field of spatial information to provide methods for extracting features. Remote sensing technology and satellite imagery have become an important tool for obtaining information in order to extract features. The presence of some obstacles such as weather conditions and the presence of clouds and shadows in satellite imagery prevents us from getting information from the surface of the earth. To solve this problem, we examined the ability of the radar images in helping to extract the urban features by optical images, especially detecting the pixels located in shadow and cloud areas. In this paper, the images of WorldView-3, ALOS-2 with single polarization and four polarization were taken into consideration and the optimally extracted features were used to classify the vegetation, building, road and soil using feature and decision level fusion. The optical features include the gray-level co-occurrence matrix (GLCM) and the radar features include GLCM for the image of single polarization, the features of the target decomposition, the separation characteristics and the main characteristics for the image with full polarization. In the classification by optical and radar features using feature level fusion, the combined accuracy of 83.96 percent was obtained and somewhat it was able to correctly identify the pixels in the shadow and cloud areas, while the classification with the optical features obtained a total accuracy of 81.02 percent. The obtained results of the decision level fusion are very low and unacceptable. The results in this paper showed that the use of the radar images along with the optical images in the feature classification using the feature level fusion improved the accuracy to some extent; and depending on the different conditions the result may be different. |
|
| Keywords: Feature Extraction, Optic and Radar Images, Feature Level Fusion |
|
|
Full-Text [PDF 1148 kb]
(418 Downloads)
|
Type of Study: Research |
Subject:
RS
Received: 2019/03/11 | Accepted: 2019/09/15 | ePublished ahead of print: 2023/06/21 | Published: 2023/07/9
|
|
|
|
|
|
|
| Send email to the article author |
|
|
