基于Dropblock卷积神经网络的分纤箱智能质检

Intelligent Quality Inspection of Fiber Separation Box Based on Dropblock Convolution Neural Network

针对传输网络中分纤箱的质检,提出一种基于Dropblock卷积神经网络的分纤箱智能质检方法.首先,通过数据初次预处理及基于Resnet的初次识别,识别出有无分纤箱;其次,对数据进行二次预处理,然后通过基于Dropblock的卷积神经网络进行二次识别,识别出分纤箱上是否有保护线和标签.实验结果表明,该方法能有效地识别出分纤箱上的保护线及标签.

Aiming at the quality inspection of fiber separation box in transmission network,an intelligent quality inspection method of fiber separation box based on Dropblock convolution neural network was proposed.Firstly,through the initial data preprocessing and initial identification based on Resnet,identify whether there is a fiber separation box.Secondly,the data is preprocessed for the second time,and then the secondary identification based on Dropblock convolution neural network is carried out to identify whether there are protection lines and labels on the fiber separation box.The experimental results show that the method can effectively identify the protection lines and labels on the fiber separation box.