Enhanced Detection and Classification of Underwater Objects using ROV and Computer Vision

Abstract

Among the various challenges in underwater exploration, the identification and classification of objects, especially metallic items, hold significant importance in diverse contexts. This paper introduces a comprehensive algorithmic framework leveraging ROVs and computer vision to detect and classify metallic objects in aquatic environments. The Experimental Design section outlines the multi-step process employed for underwater object detection using ROVs. The algorithm undergoes image enhancement, YOLOv3-based object detection, and CNN-based object classification. The dataset used for training and testing comprises a diverse set of underwater scenes with varying illumination, object sizes, and background complexities. The Results and Analysis section presents the performance evaluation of the integrated algorithm. Standard metrics for object detection, including Intersection over Union (IoU), precision, recall, and F1 score, are utilized. The algorithm demonstrates high accuracy in detecting various metallic objects. The comparative analysis of precision, recall, and F1 score across different classes further validates the algorithm's effectiveness in identifying and classifying specific objects underwater.