Boat Monitoring and Classification System utilizing infrared and visual data streams for port safety enhancement

This paper presents a prototype monitoring and classification system designed to enhance safety in port areas. The system utilizes infrared and visual data streams from four cameras to detect and classify incoming and outgoing boats, even in adverse weather conditions. Deep neural networks, specifically region-based Convolutional Neural Networks (CNNs) and the You Only Look Once (YOLO) network, are employed for boat detection and classification. The experimental hardware setup, including camera specifications and positioning, is described. The paper discusses different architectures for object detection and compares their performance. A discussion on future developments and the potential for extending the application of the developed system to other ports.