Advanced Techniques for Maritime Target Detection and Recognition Using YOLO-v7 and Sliding Window Segmentation

This paper presents an advanced methodology for maritime target identification utilising the latest YOLO-v7 object detection algorithm and Unmanned Aerial Vehicles (UAVs). The study integrates sliding window segmentation and Non-Maximum Suppression (NMS) techniques to enhance detection accuracy in complex maritime environments. Various preprocessing techniques, including image augmentation and standardisation, are applied to publicly accessible maritime databases and imagery captured by the Albatross II UAV. The proposed method demonstrates significant improvements in detection precision, leveraging deep learning techniques to address challenges posed by variable weather conditions and diverse object sizes. Experimental results indicate that the enhanced YOLO-v7 algorithm combined with sliding window segmentation increases detection accuracy by 1-2%.