Ship detection in maritime environments presents significant challenges due to factors like dramatic scale variations, complex backgrounds, and flexible viewpoints in drone-captured images, traditional object detection methods struggle with these conditions, requiring advanced deep learning techniques for effective identification and in this paper, we propose a deep learning-based ship detection system designed to overcome these challenges and the system integrates Gray-Level Co-occurrence Matrix (GLCM) and Histogram of Oriented Gradients (HOG) for feature extraction, enhancing the model's ability to differentiate ships from cluttered backgrounds, a deep neural network with Dropout layers is employed to improve generalization and reduce overfitting. The model Is trained on synthetic datasets and evaluated using key performance metrics like precision, recall, F1-score, accuracy, and ROC-AUC curves, results demonstrate high detection accuracy, effectively minimizing false positives and false negatives, additionally, threshold optimization enhances detection performance across diverse maritime conditions, to further improve efficiency, future research may incorporate Vision Transformers (ViTs) to enhance contextual understanding and reinforcement learning for adaptive detection in dynamic environments and the proposed system contributes to real-time ship monitoring and maritime security, offering a scalable solution for autonomous surveillance and vessel tracking.
Toward Real-Time Maritime Surveillance: Deep Learning and Feature Fusion for Ship Detection
28.05.2025
354113 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Online learning for ship detection in maritime surveillance
Tema Archiv | 2010
|Data Fusion Architecture for Maritime Surveillance
British Library Conference Proceedings | 2002
|Research on Real-Time Ship Detection Using Deep Learning
IEEE | 2022
|Data fusion architecture for Maritime Surveillance
IEEE | 2002
|