基于DCNN技术的无人值守变电站智能监测模型优化设计

Optimization design of Intelligent monitoring model for uattended substation based on DCNN technology

为了改善变电站监测图像峰值信噪比、结构相似性差的问题,提出基于DCNN模型的无人值守变电站监测图像超分辨率智能重建方法。通过降质退化模型分析图像劣化过程,并利用硬和软阈值结合方法去除噪声,构建8层DCNN模型结构,得到特征图,将高、低分辨率图像以图像对的形式组合,计算对应的边缘图像对,将这4个特征图用同一个稀疏表示约束,联合图像,智能重建图像。实验结果表明:重建图像的细节表现丰富,边缘信息保留完整,并且峰值信噪比为47.5 dB,结构相似性为0.97,验证了该方法具备有效性。

In order to improve the problems of peak signal-to-noise ratio and poor structural similarity of substation monitoring images,a super-resolution intelligent reconstruction method of unattended substation monitoring images based on DCNN model is proposed.The degradation model was used to analyze the process of image degradation,and the combined method of hard and soft threshold was used to remove the noise.An 8-layer DCNN model structure was constructed to obtain the feature maps.The high and low resolution images were combined in the form of image pairs,and the corresponding edge image pairs were calculated.The four feature maps were constrained by the same sparse representation,and the images were jointly reconstructed intelligently.The experimental results showed that the reconstructed images were rich in detail,the edge information was intact,the peak signal-to-noise ratio was 47.5 dB,and the structural similarity was 0.97,which verified the effectiveness of the proposed method.