Leveraging Deep Learning and Computer Vision for Accurate Maritime Vessel Activity Detection

Proposed approach is unique in that it employs computer vision techniques to classify trajectories and detect activities in mobility data. With graphically portraying vessel motions as images, proposed method, in particular, makes advantage of image classification techniques. The work presented here describes a novel method for automatically converting streaming vessel trajectory patterns into images, and for categorizing vessel activities in near-real-time by using deep learning algorithms. It addresses the challenges of Big Data volume and velocity effectively. This research used a Deep Learning CNN system to monitor trajectory (movement) patterns of sea vessels (Transportation) ships to detect their activities. Nowadays, all vehicles are fitted with sensors that detect present location & movement of vehicle or vessel & subsequently report to a centralized server, allowing us to identify vessel activities with analyzing trajectory. Applied PSO (Particle Swarm Optimization) algorithm to optimize image features & then VGG16 will get trained on PSO optimize features to get better prediction or classification accuracy. PSO algorithms can extract clear pixels & most suitable pixels from images & due to that reason deep learning algorithm accuracy will increase as its getting trained on suitable pixels. PSO algorithm will take all pixels of images as input & then it will find pixels which are more related to one & other & this related pixels will endure input to VGG16 deep learning algorithm as its trained on related pixels data so its accuracy will get increase.