Transfer Learning Based Convolutional Neural Network for Classification of Remote Sensing Images
Classification of Land cover Remote sensing images find a lot of applications including regional planning, natural resources conservation and management, agricultural monitoring etc., Presently, Convolutional Neural Networks (CNN) which are deep learning based methods are successfully employed for...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Stefan cel Mare University of Suceava
2023-11-01
|
| Series: | Advances in Electrical and Computer Engineering |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.4316/AECE.2023.04004 |
| Summary: | Classification of Land cover Remote sensing images find a lot of applications including regional planning, natural
resources conservation and management, agricultural monitoring etc., Presently, Convolutional Neural Networks (CNN)
which are deep learning based methods are successfully employed for classification problems due to its flexible
architecture and potentiality to learn new features from raw data. The motivation of the work is to implement
a robust deep learning architecture for the classification of remote sensing images using a transfer learning
approach. Deep learning requires a large amount of time if the training is initiated from scratch. Transfer
learning overcomes this drawback by using pre-trained models efficiently. In the proposed work, a transfer
learning based Convolutional Neural Network is used for the classification of remote sensing images. Three
popular pre-trained models – VGG16, ResNet50 and Densenet121 are used for feature extraction and a fully
connected layer is used for classification. Results indicate that the transfer learning based Convolutional
Neural Network with data augmentation and optimization of model parameters gives better performance compared
to training from scratch for the classification of remote sensing images. Experimental results indicate that
an improved accuracy of 95.88 percent is obtained for the proposed Transfer learning method for the remote
sensing dataset of UC-Merced. |
|---|---|
| ISSN: | 1582-7445 1844-7600 |