Deep Learning Classification of Empty and Full Container Ships in Satellite Imagery

Abstract

This chapter presents a novel approach to ship classification in optical remote sensing (ORS) imagery, focusing on the distinction between empty and full container ships. Leveraging deep cognitive modeling techniques, the study employs renowned pre-trained deep learning models, including VGG-16, VGG-19, and InceptionV3, with fine-tuning for enhanced performance. The investigation addresses the challenges posed by class imbalance through strategic data augmentation. Results demonstrate the efficacy of the proposed models, with InceptionV3 exhibiting superior performance. Evaluation metrics encompassing accuracy, precision, recall, F1-score, AUC-ROC, and AUC-PR are meticulously analyzed. These findings contribute to the advancement of ship classification methodologies in ORS imagery, with implications for maritime applications and decision-making processes. The work underscores the importance of deep cognitive modeling in addressing complex classification tasks and paves the way for future enhancements and applications in the field.