Published Paper Details:

Satellite imagery stands as a pivotal tool across diverse sectors such as agriculture, defense and finance, offering unique insights. This study focuses on the crucial objective of automating the identification of ships in satellite imagery, with a particular emphasis on a dataset obtained from the San Francisco Bay and San Pedro Bay regions. The investigation systematically evaluates the effectiveness of ANN and CNN models. The ANN model is designed to capture complex relationships within the pixel values, while the CNN model leverages its spatial hierarchies to detect intricate patterns. The study assesses the robustness of the models to atmospheric conditions, evaluates the potential benefits of ensemble methods, and investigates the feasibility of real-time ship detection applications using ANN and CNN. Furthermore, the influence of image resolution on the accuracy of both models is analyzed, and the transferability of trained models to different geographical regions is examined. The findings contribute insights into the nuances of implementing ANN and CNN models for ship detection, addressing challenges and opportunities in maritime surveillance using satellite imagery. The models achieved an exceptional accuracy of 99.37%, underscoring their efficacy in ship detection.