An Ensemble Method for Volcano Detection on Venus Using Magellan Spacecraft Radar Images

Volcanic activity on Venus using radar data from the Magellan spacecraft provides deep insights into planetary science. This research proposes an ensemble deep learning method for automated volcano detection on Venus based on radar imagery. The ensemble combines the three state-of-the-art convolutional neural networks, MobileNetV2, DenseNet161, and EfficientNetB0, which integrate with the meta-model (logistic regression) for the classification of volcano and no-volcano classes. The accuracy of the proposed ensemble method was 98.65 %, which was higher than that of the individual models MobileNetV2, DenseNet161, and EfficientNetB0, which were <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 7. 7 3 \%, 9 7. 4 4 \%, }$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 6. 4 9 \%, }$</tex> respectively. However, the ensemble method has the lowest loss of 4.73 % compared to other pre-trained models. Furthermore, the ensemble method provided superior results, where the weighted accuracy, recall, and F1-score of the ensemble method were consistently <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 0 0 \%}$</tex>. For processing large planetary datasets, this high-performance method can provide a scalable framework. Our findings might guide future missions and can show how effective ensemble learning is in advancing planetary exploration.