基于判别式深度置信网络的智能电缆隧道缺陷检测技术研究

Fault detection technology for smart cable tunnel based on discriminant deep belief network

针对智能电缆隧道缺陷检测的需求,提出了一种基于判别式深度置信网络的电缆缺陷检测技术。其通过结合神经网络与Softmax回归层模型,构建了判别式深度置信网络模型,并将智能设备采集的图像信息作为训练集进行无监督贪婪训练。同时为了进一步提高缺陷检测能力,构建了损失函数与惩罚函数,并利用梯度下降法对网络模型进行反向无监督训练。仿真结果表明,所提算法相对于现有算法能够更加准确地识别出电缆缺陷,且对于不同类型的缺陷均具有良好的适用性。

In response to the demand for defect detection in smart cable tunnels,a cable defect detection technology based on discriminative deep confidence networks is proposed.By combining the neural network and the Softmax regression model,a discriminative deep belief network model is constructed,and the image information collected by the smart device is used as the training set for unsupervised greedy training.In order to further provide defect detection capabilities,this paper constructs a loss function and a penalty function,and uses the gradient descent method to perform reverse unsupervised training of the network model.The simulation results show that the algorithm proposed in this paper can identify cable defects more accurately than existing algorithms,and has good applicability to different types of defects.