Enhanced Small Ship Detection Method for Unmanned Surface Navigation Using the Divide and Conquer Approach and ByteTrack.

Ship detection is a fundamental technology in autonomous navigation for Unmanned Surface Vehicle(USV). Deep learning-based object detection models are commonly employed for ship detection. However, they have limitations, such as reduced accuracy when detecting distant or small objects. Moreover, detection results can vary depending on the shape and background of the object. In this paper, we propose a ship detection and tracking method that utilizes the divide and conquer approach along with the tracker to address these challenges. By employing the divide and conquer approach, we enhance detection accuracy for small objects through region selection. Furthermore, ByteTrack is employed to ensure reliable detection results, even in the presence of changing object appearances. The proposed method was applied to ship images acquired from actual USV, and as a result, it shows better results than the existing object detection.