A Deep Learning Framework for Enhancing Maritime Coastal Security

The concerns of coastal security are very dynamic which depends on several factors like the neighbourhood of the country, the terrain and the marine traffic. As in the present system, only the vessels which emit Automatic Identification System (AIS) data streams can be detected and identified. Small vessels like some boats do not have AIS system setup, so they can get past without getting detected. Therefore, we propose a solution by framing a deep learning architecture namely faster region based convolutional neural network (Faster R-CNN) which accurately detects vessels in satellite images and outputs the latitude and longitude coordinates which when merged with AIS data helps identify whether the vessel is registered or not. Along with this, our paper also focuses on threat level detections from the unregistered vessels to analyze any terrorizing activity that may occur along the coastline.