| 摘要: |
| 现阶段光网络哑资源的管理多采用人工方式进行监控管理,而引入电子标签的智能光分配网络（Optical Distribution Network,ODN)因成本高、改造实施困难导致运营商无法有效提高ODN网络质量。通过引入二维码标签及标签分别对分光器及尾纤进行标识，利用深度学习技术，采用Faster-RCNN网络对哑资源中尾纤、尾纤标签、分光器、二维码等信息进行统一识别管理，mAP(Mean Average Precision)达82.96%，满足实际需求，实现了哑资源智能管理，降低了管理与维护成本。 |
| 关键词: 光分配网络（ODN) 哑资源 智能管理 二维码标签 图像识别 深度学习 |
| DOI： |
|
| 基金项目: |
|
| A deep learning intelligent management system for optical network dumb resources |
| ZHANG Gaoyi,ZHANG Jun,LIU Wei |
| (Postdoctoral Scientific Research Workstation,China Mobile Communications Group Sichuan Co.,Ltd.,Chengdu 610041,China;School of Mathematics and Computer Science,Guangdong Ocean University,Zhanjiang 524003,China;Huaxin Consulting Design Research Institute Co.,Ltd.,Hangzhou 310052,China) |
| Abstract: |
| At present,manual monitoring management is usually adopted for the management of optical network dumb resources.The high cost of introducing electronic tag in intelligent optical distribution network(ODN) makes it difficult for operators to effectively improve the quality of ODN network.The quick response(QR) code tag and normal tag are introduced to identify splitter and pigtail,and deep learning technology and faster-RCNN network are used to identify and manage the pigtail,pigtail label,splitter,QR code and other information in dumb resources.The mean average precision(mAP) is 82.96%,which meets the actual needs,thus realizing intelligent management and reducing the management and maintenance costs. |
| Key words: optical distribution network(ODN) dumb resource intelligent management QR code tag image recognition deep learning |
