Novel Approach to Underwater Object Detection Using Sonar Sensors for Autonomous Underwater Vehicles (AUVs)

Object detection in sonar images is a vital component of Autonomous Underwater Vehicles (AUVs), serving underwater exploration, surveillance, and environmental monitoring applications. Despite the unique challenges posed by sonar data, including noise, limited resolution, and varying acoustic conditions, this paper presents a comparative study of YOLOv8, a state-of-the-art deep learning model optimized for object detection in sonar images. The study outlines the adaptations made to YOLOv8 to enhance its suitability for sonar data, introduces a meticulously curated sonar image dataset, and details data preprocessing techniques. Rigorous experiments evaluate the performance of YOLOv8, considering key metrics like precision, recall, F1-score, and mean average precision (mAP). Our results reveal valuable insights into the suitability of YOLO-based models for sonar image object detection in AUVs. We compare the detection accuracy, speed, and robustness of YOLOv8 in underwater environments. Furthermore, we discuss the practical implications of our findings and highlight potential areas for future research. This research contributes to the advancement of autonomous underwater systems by enhancing their ability to detect and identify objects of interest in challenging sonar imagery. The findings presented in this paper provide valuable guidance for selecting and optimizing deep learning models for real-world underwater applications.