Seafloor Sediment Detection with Sidescan Sonar Image Based on YOLO11 and YOLO26

ABSTRACT

The study of seafloor geomorphology is the core foundation for decoding marine geological evolution. It prevents and controls marine engineering hazards, and supports marine resource development and ecological protection. AUVs equipped with side‐scan sonar (SSS) are the core tool for seafloor exploration. However, traditional manual interpretation of SSS images has obvious limitations. It has low efficiency and a high false detection rate in complex terrain. This paper focuses on the demand of seafloor sediment detection using SSS images. It systematically compares the architectural differences and comprehensive performance of YOLO11 and YOLO26. This study uses the measured SSS dataset from the Taiwan Strait. The dataset is processed by data augmentation. The experimental results show that YOLO26 has significant advantages in precision, comprehensive positioning accuracy and inference speed. YOLO11 has a higher target recall rate. The two models form differentiated adaptation for different application scenarios. This study provides experimental basis for the selection of detection models in seafloor exploration scenarios.