Boat Detection and Classification Framework for Safety Improvement in Port Areas

The detection and classification of boats is an important issue that needs to be addressed to improve safety in harbour areas. This paper presents a monitoring system based on multiple infrared and visual cameras installed in a tourist harbour in Augusta. Deep neural network models based on YOLO networks were trained to locate and classify incoming and outgoing boats in the harbour area using streaming data from visual and infrared cameras. The performance of the monitoring system and the use of visual data to create a risk indicator useful for harbour security management are discussed using the implemented framework in different scenarios identified from the developed dataset collected by the monitoring system.