基于图像特征提取与BP神经网络的绝缘子憎水性识别方法

发布时间：2018-05-17 13:11

本文选题：绝缘子 + 憎水性　；参考：《高电压技术》2014年05期

【摘要】：为了方便、快捷、准确地识别运行中绝缘子表面的憎水性等级,提出了基于图像特征提取与BP神经网络的绝缘子憎水性识别方法。首先,利用同态滤波和直方图均衡对憎水性图像进行增强预处理,然后利用改进的Canny算子提取了图像中水珠(或水迹)的边缘,并利用数学形态学对其进行修正得到最终的分割图像;提取图像中与憎水性相关的4个特征量,最后建立了基于BP神经网络的憎水性识别模型,并对测试样本的憎水性等级进行了识别。实验结果表明,该方法克服了人为因素的影响,能够有效识别绝缘子7种典型憎水性等级,总识别率达90%,为在线检测绝缘子憎水性提供一种新的有效方法。
[Abstract]:In order to identify the hydrophobicity of insulator surface conveniently, quickly and accurately, a hydrophobicity recognition method based on image feature extraction and BP neural network is proposed. Firstly, homomorphic filtering and histogram equalization are used to enhance the preprocessing of hydrophobic image, and then the edge of water droplets (or traces) in the image is extracted by using the improved Canny operator. Finally, the hydrophobicity recognition model based on BP neural network is established, and the final segmentation image is obtained by modifying it with mathematical morphology, and the four characteristic quantities related to hydrophobicity are extracted from the image, and the hydrophobicity recognition model based on BP neural network is established. The hydrophobicity grade of the test sample was identified. The experimental results show that this method overcomes the influence of human factors and can effectively identify 7 typical hydrophobicity grades of insulators with a total recognition rate of 90. It provides a new and effective method for on-line detection of insulator hydrophobicity.
【作者单位】：华北电力大学河北省输变电设备安全防御重点实验室;
【基金】：中央高校基本科研业务费专项资金(13MS71)~~
【分类号】：TP391.41;TM216

【参考文献】