|
|||||
|
|
| 基于深度学习的水面无人船前方船只图像识别方法 | |
| 引用本文： | 王贵槐,谢朔,初秀民,洛天骄. 基于深度学习的水面无人船前方船只图像识别方法[J]. 船舶工程, 2018, 40(4): 19-22 |
| 作者姓名： | 王贵槐 谢朔 初秀民 洛天骄 |
| 作者单位： | 武汉交通职业学院,武汉,430065;武汉理工大学国家水运安全工程技术研究中心,武汉430063;武汉理工大学能源与动力工程学院;武汉理工大学能源与动力工程学院;武汉理工大学物流工程学院,武汉,430063 |
| 基金项目： | 武汉市科技计划项目（2017010201010132） |
| 摘 要： | 建立基于图像识别系统的水面无人船感知平台，采集内河船舶图片数据库建立船只检测单层多尺度深度学习（Single Shot Multibox Detector，SSD）框架，通过使用预训练模型参数调优并微调分类框架实现较高的内河船舶检测准确度。试验结果表明，不同天气状况下的识别算法的查全率和查准率均能保持在70%以上 |
| 关 键 词： | 深度学习 水面无人船 图像识别 单次多重检测器 |
| 收稿时间： | 2017-11-15 |
| 修稿时间： | 2018-04-25 |
Research on image recognition method of ships in front of unmanned surface vessel based on deep learning |
|
| wang guihuai,chu xiumin and luo tianjiao. Research on image recognition method of ships in front of unmanned surface vessel based on deep learning[J]. Ship Engineering, 2018, 40(4): 19-22 | |
| Authors: | wang guihuai chu xiumin luo tianjiao |
| Abstract: | Obstacle perception is the basis of autonomous navigation of USV (unmanned surface vessel). This paper established the USV perception platform based on the image recognition system. Firstly, a SSD (single shot multibox detector) deep learning framework is built to detect the ship by the collected inland ship picture database. Secondly, with the pre training model parameters tuning and fine-tuning of the classification framework, a high target recognition accuracy of inland ships is achieved. Finally, experimental results show that the recall and precision of the proposed recognition algorithm can almost reach more than 70% under different weather conditions. |
| Keywords: | USV target recognition deep learning SSD |
| 本文献已被 万方数据 等数据库收录！ | |
| 点击此处可从《船舶工程》浏览原始摘要信息 | |
| 点击此处可从《船舶工程》下载全文 | |