Optimized YOLOv7 Deployment for Underwater Surveillance Drones

Abstract

Deep learning has gained significant attention in recent years for its potential in categorising underwater photographs to identify various objects like fish, plankton, coral reefs, sea grass, submarines, alien objects and the movements of a deep - sea diver for subaquatic surveillance. Accurate classification is essential for the surveillance of the aqua condition and purity of water/sea bodies, as well as the preservation of endangered species living inside it. Additionally, it has importance in the fields of maritime affairs, defence and security. Thus, we have proposed a system which utilizes a deep convolutional neural network (VGG16), a type of deep learning technology, to recognize the image in order to provide underwater monitoring with great convenience. Grayscale and white balance techniques have been employed to reduce complexity and enhance the quality of underwater images before using Deep CNN. Ultimately, the experimental study confirms the successful identification of the acquired underwater images.