基于改进GoogLeNet模型的电气设备红外图像分割与识别

Infrared image segmentation and recognition of electrical equipment based on improved GoogLeNet model

针对红外电气设备图像对比度低和信噪比低导致图像识别率不高的问题,提出一种改进的红外图像分割与识别方法。首先通过K均值聚类结合区域生长算法实现对图像的预处理,通过GoogLeNet网络对图像特征进行提取;其次在图像识别阶段,为提高识别率,引入超参数构建联合损失函数,以此改进GoogLeNet网络训练的识别率;最后搭建深度学习算法实验环境,并以实验室搜集的电气设备红外故障图像为例对图像进行识别。结果表明,所提方法可有效提取图像特征,并且识别率要明显高于improve_cnn、VggNet、AlexNet 3种模型。

In view of the low contrast and low signal-to-noise ratio(SNR)of electrical equipment images,an improved infrared image segmentation and recognition method is proposed.Firstly,the image is preprocessed by K-means clustering combined with region growing algorithm,and then the image features are extracted by GoogLeNet network;In the image recognition stage.In order to improve the recognition rate,a joint loss function is introduced to improve the recognition rate of GoogLeNet network training.Finally,the experimental environment of deep learning algorithm is set up,and the infrared fault image of electrical equipment collected in the laboratory is taken as an example to identify the image.The results show that this method can effectively improve the extraction of image features,and the recognition rate is significantly higher than improve_cnn,VggNet and AlexNet.