Classification of Marine Vessels using Deep Learning Models based on SAR Images

In recent years, concern has been increased on marine resources due to illegal fishing activities and the degradation of ecosystems. Earth's surface is roughly covered by 70% of water. Few countries use these water bodies as a buffer zone to safeguard their countries from threats. They also serve as a mode of transportation as well as shelter for a huge number of marine lives. In consequence, tracking various types of marine vessels and their activities demand to focus on new research and extracting the underlying information. The radar images captured by satellites are extremely useful in solving these problems. In this paper, various types of Artificial Neural Networks (ANNs) have been used to find out various patterns of marine vessels through SAR-captured images. Specifically, a few classification algorithms like CNN, ResNet50, and VGG-16 have been applied to classify maritime vessels. It is expected that using the classification model with the highest accuracy can be utilized to predict the vessel’s pattern which would be helpful for concerned authorities to monitor the marine traffic and restrict the vessels from illegal entry into water bodies. It is also concluded that CNN is the best model providing an accuracy of 99.25% by defeating the ResNet50 and VGG-16 models.