摘要
海上舰船目标分类检测技术可广泛用于海事监管、船只救援、打击非法犯罪等活动,具有重要应用价值。本文选取高分辨率光学遥感影像进行数据筛选及预处理,建立了含有2.4万艘不同类型舰船的分类检测数据集。依据中华人民共和国海事局《船舶登记工作规程》中的舰船分类规则,结合遥感影像实际情况,建立了基于遥感影像的海上舰船分类体系。通过搭建深度学习训练平台,使用YOLOv3神经网络算法对舰船进行分类检测。在测试集上对训练完成的模型进行验证,舰船分类检测结果的召回率达到91%以上,准确率达到95%以上。在GPU加速的情况下,可达30 fps以上的检测速率,使得该模型在具备鲁棒性和准确性的同时也具备实时分类检测的能力。
Ship detection and classification technology has important application value.It can be widely used in maritime supervision,ship rescue,and combating illegal crimes.High-resolution optical remote sensing images were selected for data screening and pre-processing,and a marine ship classification detection data set containing more than 24,000 ships of different types was established.According to the ship classification rules in the"Regulations on Ship Registration"of the Maritime Safety Administration of the People’s Republic of China,combined with the actual situation of remote sensing images,a basic classification system for marine vessels based on remote sensing images was established.A deep learning training platform was built,and the ship was classified and detected using the YOLOv3 neural network algorithm.The model completed by the training is verified on the test set.The recall rate of the ship classification test results is over 91%,and the accuracy rate is over 95%.In the case of GPU acceleration,the detection rate above 50 fps can be achieved,which makes the model robust and accurate,as well as real-time classification detection.
作者
王浩君
周斌
潘玉良
Wang Haojun;Zhou Bin;Pan Yuliang(Institute of Remote Sensing and Earth,Hangzhou Normal University,Hangzhou 310000,China)
出处
《科技通报》
2020年第3期43-48,58,共7页
Bulletin of Science and Technology
作者简介
王浩君(1993-),男,硕士研究生,研究方向为遥感目标识别。E-mail:505832123@qq.com。
