An AI-Powered Autonomous Underwater System for Sea Exploration and Scientific Research

An AI-Powered Autonomous Underwater System for Sea Exploration and Scientific Research 2nd Mariam Al Nasseri College of Technological Innovation Zayed University Abu Dhabi, United Arab Emirates 202111707@zu.ae.ac 1st Hamad Almazrouei College of Technological Innovation Zayed University Abu Dhabi, United Arab Emirates 201912368@zu.ae.ac 3rd Maha Alzaabi College of Technological Innovation Zayed University Abu Dhabi, United Arab Emirates 202104533@zu.ae.ac Abstract—Traditional sea exploration faces significant chal- lenges due to extreme conditions, limited visibility, and high costs, resulting in vast unexplored ocean regions. This paper presents an innovative AI-powered Autonomous Underwater Vehicle (AUV) system designed to overcome these limitations by automating underwater object detection, analysis, and reporting. The system integrates YOLOv12 Nano for real-time object detection, a Convolutional Neural Network (CNN) (ResNet50) for feature extraction, Principal Component Analysis (PCA) for dimensionality reduction, and K-Means++ clustering for grouping marine objects based on visual characteristics. Fur- thermore, a Large Language Model (LLM) (GPT-4o Mini) is employed to generate structured reports and summaries of underwater findings, enhancing data interpretation. The system was trained and evaluated on a combined dataset of over 55,000 images from the DeepFish and OzFish datasets, capturing diverse Australian marine environments. Experimental results demonstrate the system’s capability to detect marine objects with a mAP@0.5 of 0.512, a precision of 0.535, and a recall of 0.438. The integration of PCA effectively reduced feature dimensionality while preserving 98% variance, facilitating K-Means clustering which successfully grouped detected objects based on visual similarities. The LLM integration proved effective in generating insightful summaries of detections and clusters, supported by location data. This integrated approach significantly reduces the risks associated with human diving, increases mission efficiency, and enhances the speed and depth of underwater data analysis, paving the way for more effective scientific research and discovery in challenging marine environments. Index Terms—Autonomous Underwater Vehicles (AUVs), Ob- ject Detection, Deep Learning, Underwater Exploration, K- Means Clustering I. INTRODUCTION A. Background The sea remains one of the most mysterious and least explored regions of our planet. To begin with, extreme con- ditions, including high pressure, poor visibility, and unpre- dictable underwater landscapes, pose significant challenges to traditional exploration methods, making them dangerous, costly, and limited in effectiveness. In fact, despite covering over 70% of Earth’s surface, only about 5% of the ocean has been fully explored, underscoring the vast unknowns that persist due to these challenging conditions [1]. In addition, human divers face substantial risks and data collection is often slow and incomplete. For example, typical dive durations are limited to less than an hour (approximately 50 minutes) due to air supply constraints, and marine data collected manually can take weeks to analyze and validate, significantly slowing scientific progress [2]. However, recent advancements in Arti- ficial Intelligence (AI) and autonomous systems are creating new opportunities. As a result, by combining Computer Vision, Machine Learning, and Automated Reporting, sea exploration can now become safer, faster, and more in- sightful. Therefore, AI-powered AUVs equipped with real- time image processing and autonomous navigation are already demonstrating superior performance in identifying marine species and environmental anomalies compared to traditional methods [3]. B. Project Overview This project presents an AI-powered Autonomous Un- derwater Vehicle (AUV) System designed to transform sea exploration. The system integrates YOLOv12 (You Only Look Once) for real-time object detection with a Large Language Model (LLM) that generates structured reports and summaries of underwater findings, focusing on detected objects and clusters. It is capable of detecting and assigning marine objects to pre-defined clusters, including both known and unknown types, and automatically produces detailed reports to support scientific research, environmental monitoring, and future ocean discoveries. Furthermore, K-Means clustering is applied to analyze patterns in marine biodiversity, enabling improved classification and ecosystem understanding. C. Scope By deploying this autonomous system, the project aims to reduce the risks associated with human diving, increase the efficiency and speed of underwater missions, and overcome environmental challenges such as low visibility and physical obstacles. Autonomous systems are increasingly replacing human divers in hazardous environments, improving safety while significantly increasing the volume and speed of data collection [4]. Energy optimization is a

This content is AI-processed based on open access ArXiv data.