MAGNAT: Maritime Management Ensemble Learning System

The classification of ships using remotely sensed data is increasingly vital for maritime security, environmental monitoring, and commercial applications. Advances in satellite imagery and remote sensing technologies have heightened the need for effective ship classification algorithms. A key challenge lies in the variability of ship appearances due to size, shape, orientation, and environmental factors. Ensemble learning, which combines multiple classifiers for improved accuracy, has shown promise in this area. However, this method faces issues like increased memory and time complexity, and the influence of lower-quality models, highlighting the need for model pruning. This study presents a new approach to maritime management through the introduction of an ensemble learning system. The MAGNAT (Maritime mAnaGement eNsemble leArning sysTem) proposal and novel contribution in this paper represent an intelligent ensemble solution that significantly enhances ship classification in the maritime domain through the adoption of ensemble learning techniques. Incorporating various Convolutional Neural Network (CNN) models into the training process, a novel aggregation strategy is devised to eliminate redundant models, retaining only those that actively enhance the learning process. Extensive experiments utilizing a renowned ship dataset were conducted to validate the effectiveness of the framework, ultimately demonstrating its superiority over conventional ensemble solutions.