基于红外与紫外图像信息融合的绝缘子污秽状态识别

Recognition of Contamination Grades of Insulators Based on IR and UV Image Information Fusion

为了实现绝缘子污秽状态的非接触检测,提出了一种基于红外与紫外图像信息决策级融合的污秽等级识别方法。分别计算不同污秽等级绝缘子红外与紫外图像特征,根据Fisher准则进行特征选择,得到可以有效表征污秽状态的特征量,为了提高分类器的运算速度和准确性,利用核主元分析(KPCA)进行特征提取,分别得到红外与紫外特征的三维核主元向量,使用径向基神经网络(RBFNN)分别进行污秽等级识别,利用D-S证据理论对识别结果进行决策级融合,实现绝缘子污秽等级的识别。实验结果表明,该方法的正确率显著优于单独使用红外或紫外特征进行识别,为绝缘子污秽状态的非接触检测提供了新的方法。

In order to realize the non-contact measurement of insulator pollution severity, a method based on decision level fusion of IR and UV image information is proposed. Features of IR and UV images are calculated separately. Meanwhile, feature selection based on Fisher criterion is carried out to obtain features which have the ability to represent the contamination grades efficiently. In order to improve the calculation speed and precision of classifier, Kernel principal component analysis(KPCA) is adopted to extract three-dimensional Kernel principal features of IR and UV images. Radial basis function neural network(RBFNN) is used to identify the contamination grades using IR and UV features separately. And then, D-S theory is adopted to achieve the decision fusion and realize the high accuracy identification of contamination grades. Results of the experiments indicate that the precision of proposed method is significantly superior to recognition using IR or UV features separately. This paper provides a new method for the prevention of pollution flashover.