基于分布函数PNN的绝缘子表面污秽度检测

Insulator Surface Contamination Detection Based on Distribution Function Probabilistic Neural Network

以准确检测绝缘子表面污秽度为目的,设计基于分布函数概率神经网络的绝缘子表面污秽度检测方法。使用红外图像采集绝缘子表面信息,并完成预处理。以数学形态学和小波融合的方式,去除图像噪声信息后,将滤波后图像的直方图波谷灰度值设成分割阈值,分割获取绝缘子盘面目标图像;通过基于最小二乘拟合盘面边缘椭圆方程的特征提取方法,提取分割后绝缘子盘面目标图像的径向温度分布特征后,输入基于α稳定分布函数概率神经网络的污秽级别检测模型,通过α稳定分布函数,计算径向温度分布特征与各个污秽级别的匹配概率,完成绝缘子表面污秽度检测。实验结果显示:所提方法对绝缘子表面污秽度的检测结果符合实际,对4种污秽级别的检测结果均准确。

In order to accurately detect the contamination degree of insulator surface,a method of insulator surface contamination degree detection based on distribution function probabilistic neural network is designed.The infrared image is used to collect the insulator surface information and complete the preprocessing.In the way of mathematical morphology and wavelet fusion,after removing the image noise information,the gray value of the histogram trough of the filtered image is set as the segmentation threshold,and the target image of the insulator disk surface is segmented.Through the feature extraction method based on the least square fitting of the elliptic equation of the disc edge,the radial temperature distribution feature of the segmented insulator disc surface target image is extracted.Thenαcontamination level detection model of the stable distribution function probabilistic neural network is input throughαstably distributed function to calculate the matching probability between the radial temperature distribution characteristics and each contamination level,with the insulator surface pollution detection completed.The experimental results show that the proposed method is consistent with the actual results of insulator surface contamination detection,and that the detection results of the four pollution levels are accurate.