Vessel Detection Based on Deep Learning Approach

Abstract

An effective monitoring system to observe vessel activity is essential to provide accurate vessel position information regarding vessel activity and movement at all times. Triggered to support the current VMS and AIS monitoring systems, Vessels monitoring by applying object detection methods to find all objects of interest in an image has a chance to be implemented. This study presents a deep learning approach for processing remote sensing images to detect the presence of vessels utilizing the Faster R-CNN network as a backbone, with the extractor feature modified using the inception-v2 network. Our experiments reveal that our method yields promising results in reasonable accuracy in detecting and identifying vessels images. It achieves an accuracy of 94.4% and 0.971 for the F1Score.