基于深度学习的光学遥感影像舰船检测算法对比分析

舰船目标检测是进行海洋环境监管，保障海上权益的重要手段。基于深度学习的目标检测算法能在复杂环境下保持良好性能，为测试不同深度学习目标检测算法在舰船检测中的效果，本文构建了一个包含3893张图像的数据集，涵盖了复杂背景下不同类型的舰船，基于此数据集分别采用Faster RCNN、SSD、RetinaNet、YOLOv3、YOLOv4算法进行实验，结果表明，YOLOv4 、YOLOv3、RetinaNet、Faster RCNN平均精度均在83%以上，其中YOLOv4最高达到91.77%，Faster RCNN误检较多，而SSD平均精度最低，只有79.23%，总的舰船检测数目偏少。将5种模型训练结果在高分二号影像上进行测试，得到较好的检测效果，对舰船检测未来理论研究的开展具有一定的指导意义。

Comparative analysis of ship detection algorithms based on deep learning in optical remote sensing images

Ship target detection is an important means to supervise marine environment and protect maritime rights and interests.The target detection algorithm based on deep learning can maintain good performance in complex environment.This paper constructs a data set containing 3893 images, covering different types of ships in complex background. Based on this data set,Faster RCNN,SSD,RetinaNet,YOLOv3 and YOLOv4 algorithms are used to experiment.The results show that the mean average precision of YOLOv4,YOLOv3,RetinaNet and Faster RCNN are all above 83%.Among them,YOLOv4 reached 91.77%,and Faster RCNN has more false detections,while SSD has the lowest mean average precision,only 79.23%,and the total number of ship detections was less.The training results of the five models are tested on GF-2 image,and good detection results are obtained,which has certain guiding significance for the future theoretical research of ship detection.