摘要
为了有效的检测接触网中污秽程度较高的绝缘子,提出一种基于纹理特征和可能性均值聚类的异常检测方法,为绝缘子清洗工作的开展,提供必要的参考和前提。首先利用最大类间方差法对图像进行分割得到绝缘子的盘面区域,然后使用灰度共生矩阵计算出纹理空间的特征向量,进而使用主成分分析算法对特征向量进行融合与降维,最后通过可能性均值聚类算法实现接触网绝缘子污秽程度的异常检测。实验表明,该方法能有效的检测出污秽程度异常的接触网绝缘子。
In order to efficiently detect the insulator with high contamination in the contact net,a method based on texture feature and PCM(Possibilistic C-Means)is proposed,to provide a necessary reference for the work of insulator cleaning.Firstly,the disk area of insulator were segmented out of the segmented images by Otsu method;secondly,texture feature were calculated by GLCM(Gray Level Co-occurrence Matrix);then,the feature vectors are fused and reduced by PCA(Principal Component Analysis)algorithm;finally,the anomaly detection of the insulator contamination degree of the contact net is realized by the PCM.The experimental results show that the method can effectively detect the contamination degree of the contact net insulator.
作者
党帅涛
柯坚
吴文海
王奇
DANG Shuaitao;KE Jian;WU Wenhai;WANG Qi(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)
出处
《电瓷避雷器》
CAS
北大核心
2019年第2期197-201,共5页
Insulators and Surge Arresters
