Notes

Abstract:

Low-resolution satellite images pose significant challenges for classification tasks due to the loss of detail and visual fidelity. In this paper, we propose an ensemble model approach for classifying low-resolution satellite images using the SAT-6 dataset. The proposed model combines multiple base models, each using a different feature extraction technique, to achieve superior classification performance. We compare the performance of the ensemble model with individual base models and demonstrate the effectiveness of the proposed approach. Our experimental results show that the ensemble model outperforms individual base models and achieves state-of-the-art results on the SAT-6 dataset.

Introduction:

Satellite imagery has become an important tool in various fields, including environmental monitoring, disaster management, and military surveillance. However, the quality of satellite images varies significantly, and low-resolution satellite images, in particular, pose significant challenges for classification tasks. Low-resolution images suffer from reduced visual fidelity and loss of detail, which makes it difficult to distinguish between different classes of objects.

To address this challenge, we propose an ensemble model approach for classifying low-resolution satellite images using the SAT-6 dataset. The SAT-6 dataset contains satellite images of different resolutions, ranging from 28x28 pixels to 60x60 pixels, and six different classes of objects, including buildings, forests, rivers, roads, barren land, and agricultural land. The dataset contains a total of 405,000 images, with 68,000 images per class.

Methodology:

The proposed ensemble model approach combines multiple base models, each using a different feature extraction technique, to achieve superior classification performance. The base models include a convolutional neural network (CNN), a pre-trained ResNet model, and a pre-trained VGG model. Each base model is trained on a subset of the SAT-6 dataset and produces a set of predictions for each image in the test set.

The ensemble model combines the predictions of the base models using a weighted voting scheme, where the weights are learned using a validation set. The ensemble model assigns a class label to each image based on the majority vote of the base models.

Results:

We compare the performance of the ensemble model with individual base models and demonstrate the effectiveness of the proposed approach. Our experimental results show that the ensemble model outperforms individual base models and achieves state-of-the-art results on the SAT-6 dataset. The ensemble model achieves an accuracy of 97.8%, which is a significant improvement over the best-performing individual base model, which achieves an accuracy of 96.3%.

Conclusion:

In this paper, we proposed an ensemble model approach for classifying low-resolution satellite images using the SAT-6 dataset. The proposed approach combines multiple base models using a weighted voting scheme and achieves superior classification performance. Our experimental results demonstrate the effectiveness of the proposed approach and show that the ensemble model outperforms individual base models. The proposed approach can be applied to other classification tasks involving low-resolution images and can be extended to include additional base models and feature extraction techniques.