| 摘要: |
| 光交箱防尘帽的检测对于通信网络的正常运行具有重要作用。提出了一种基于改进Faster RCNN(Region Convolutional Neural Network)的通信网光交箱防尘帽智能检测方法。首先，对输入图片进行去噪等预处理，通过残差网络(Residual Netwok,ResNet)进行特征提取，并通过区域生成网络(Region Proposal Network,RPN）初步识别出候选区域，然后经过RolAlign进行池化处理，最后经过特征金字塔网络(Feature Pyramid Network,FPN）对光交箱防尘帽进行二次识别。将该方法应用到光交箱防尘帽缺失的智能检测中，取得了很好的效果。 |
| 关键词: 通信网 光交箱 防尘帽检测 深度学习 卷积神经网络(CNN) |
| DOI：10.20079/j.issn.1001-893x.221102005 |
|
| 基金项目:重庆市自然科学基金项目（cstc2020jscx msxmX0147） |
|
| An intelligent detection method for dust cap of communication network optical delivery box based on improved Faster RCNN |
| ZHU Hui,CHEN Jian,YUAN Jianhang |
| (China Mobile Group Sichuan Co.,Ltd.，Chengdu 610041,China;School of Computer and Information Science,Southwest University,Chongqing 400700,China;NetThink Technology Co.,Ltd.,Chengdu 610093,China) |
| Abstract: |
| The detection of dust cap of optical delivery box plays an important role in the normal operation of communication network.In this paper,an intelligent detection method for dust cap of communication network optical delivery box based on improved Faster Region Convolutional Neural Network(RCNN) is presented.Firstly,the input image is preprocessed,such as noise reduction.Secondly,feature extraction is carried out through residual network(ResNet),and the candidate regions are initially identified by Region Proposal Network(RPN).Then they are pooled by RolAlign.Finally,the dust caps of the optical delivery box are identified again by Feature Pyramid Network(FPN).The method is applied in the intelligent detection of absence of dust caps of the optical delivery box and excellent results are achieved. |
| Key words: communication network optical delivery box dust cap detection deep leaning convolutional neural network(CNN) |
