基于图像处理与RBF神经网络的绝缘子憎水性识别方法的研究

Identification of Insulator Hydrophobicity Based on Image Processing and RBF Neural Network

为了快捷准确地识别复合绝缘子的憎水性等级,提出了基于图像处理与RBF神经网络的绝缘子憎水性识别方法。首先,对图像进行直方图均衡增强、自适应中值滤波处理;然后,利用二维Ostu阈值法对图像进行分割;最后提取4个与绝缘子憎水性相关的特征量,以这4个特征量作为输入向量,以相应的憎水性等级作为输出向量,通过训练得到优化的RBF神经网络识别模型,并用于绝缘子憎水性等级的识别。试验结果表明,该方法能够准确识别绝缘子的憎水性等级,总识别率高达90%,准确度达到了实际应用的要求,为在线检测绝缘子憎水性奠定了基础。

To identify hydrophobic level of a composite insulator quickly and accurately,an identification method of insulator＇ s hydrophobic level is presented based on image processing and RBF neural network. First, histogram equalization enhancement and adaptive median filter are used to process image; Then the image is segmented by two-dimensional Otsu threshold; Finally,four features associated with hydrophobicity are extracted from the image to form the input vector, and the hydrophobic level are taken as the output vector, thus an optimal RBF neural network is constructed via training for identifying insulator＇s hydrophobic level. Testing results show that the proposed method can accurately identify hydrophobic level of insulators with total recognition rate of 90% and satisfactory accuracy. This hydrophobic level identification method may be applied to on-line detection of insulator hydrophobicity.