Seybold Report ISSN: 1533-9211
U. Deepika
Assistant Professor, Department of Electronics and Communication Engineering, Sridevi Women’s Engineering College, Hyderabad, India, deepika.upadrasta@gmail.com
G. Srishma
U.G Student, Department of Electronics and Communication Engineering, Sridevi Women’s Engineering College, Hyderabad, India, srishmareddi57@gmail.com
K. Ruchitha Chary
U.G Student, Department of Electronics and Communication Engineering, Sridevi Women’s Engineering College, Hyderabad, India, buduruchandana@gmail.com
B. Chandana
U.G Student, Department of Electronics and Communication Engineering, Sridevi Women’s Engineering College, Hyderabad, India, ruchithachary2@gmail.com
Vol 17, No 07 ( 2022 ) | DOI: 10.5281/zenodo.6879594 | Licensing: CC 4.0 | Pg no: 223-233 | Published on: 25-07-2022
Abstract
The main objective of remote sensing image scene classification is to know semantics of land covers. As CNN algorithm is used for feature representation so CNN based classification methods has been proposed for RS images. The existing system is good to capture images, which gives global information but sometimes it fails to optimise the local features. To overcome the limitation we proposed a new CNN model, for getting the local features comprehensively. It is possible due to the dual-branch structure, the input data are the image pairs that are obtained by the spatial rotation.
Considering the influence of the spatial rotation and the similarities between RS images, we developed a model to unify the salient regions and impact the RS images from the same/different semantic categories. Finally, the classification results can be obtained using the learned features. The popular RS scene datasets are selected to validate our CNN. Compared with some existing networks, the proposed method can achieve better performance. The results obtained are effective for the RS image scene classification.
Keywords:
Remote sensing, Scene Classification, CNN