基于图像分析和深度学习的复合绝缘子憎水性分级

为了更加方便快捷地检测大量复合绝缘子憎水性等级，提出一种基于图像分析和深度学习的复合绝缘子憎水性分级方法。首先为提高图像对比度，对复合绝缘子憎水性图像进行灰度化和图像增强处理；其次利用图像分析技术和U-Net网络提取水珠轮廓，得到水珠轮廓图像；接着引入深度卷积神经网络，将这些水珠轮廓图作为神经网络的输入，以相应的憎水性等级作为输出向量，训练网络得到分级模型；最后将分级模型用于憎水性分级，得到分级结果。实验结果表明:该方法的分级结果已达到实际应用要求，水珠轮廓提取的精度达到了92.96%，分级准确率达到了90.2%，预测一幅图像的憎水性等级平均耗时0.1 s。

Hydrophobicity Classification of Composite Insulators Based on Image Analysis and Deep Learning

To detect the hydrophobicity level of a large number of composite insulators more conveniently and quickly， a method for grading the hydrophobicity of composite insulators based on image analysis and deep learning was proposed. First， to improve the image contrast， the water droplet images of composite insulators were gray-scaled and image-enchanced. Then the image analysis technology and U-Net network were used to extract the contour of the water droplet to obtain the contour image of the water droplets. After that， the deep convolutional neural network was introduced， while these water droplets contours acted as the input of the neural network and the corresponding hydrophobic level as the output vectors， and the network was trained to obtain the classification model. Finally， the classification model was used for the hydrophobicity classification to obtain the classification results. The experimental results showed that the classification results of this method have reached the practical application requirements， the accuracy of water droplets contour extraction and the grading accuracy rate reach 92.96% and 90.2%， respectively； and it takes an average of 0.1 s to predict the hydrophobicity of an image.