YOLOv8-based ship target rotation detection method

The invention provides a ship target rotation detection method based on YOLOv8. The previous research is mainly a mode of carrying out horizontal ship detection from a maritime image sequence. As some background pixels may be wrongly identified as ship contours, this may result in unsatisfactory ship detection performance. In order to solve the problem, a new YOLO-based rotation model (RTM) is provided, and the ship is accurately and quickly detected from the maritime image by considering the rotation angle of the ship. The provided RTM model detects the inclined ship from the image through a BiFPN feature fusion method, an attention mechanism and a loss function GIoU, an innovative method is provided for efficient and accurate detection of ocean ships, and the method has important practical application value and can be applied to real-time ship detection tasks. 本发明提出了一种基于YOLOv8的舰船目标旋转检测方法，以往的研究主要是从海事图像序列中进行水平船舶检测的方式。由于一些背景像素可能被错误地识别为船舶轮廓，这可能导致船舶检测性能不满意。为了解决这个问题，我们提出了一种新的基于YOLO的旋转模型(RTM)，通过考虑船舶旋转角度，从海事图像中准确而快