基于SVM多特征融合的绝缘子缺陷检测算法研究

Research on Insulator Defect Detection Algorithm Based on SVM Multi-feature Fusion

绝缘子是输电线路中常用部件,每年因绝缘子故障引起的事故超过电网故障的一半,这严重危害着线路的使用和运行寿命。因此,及时对绝缘子故障进行排查是十分重要的。我们提出了一种基于SVM多特征融合的绝缘子缺陷检测算法,该算法首先对原始图像进行预处理,得到暗通道特征层和灰度图像,然后在此基础上提取出HSI特征层、细节特征层、统计特征层和纹理特征层,将上述特征层叠加成一个三维的特征矩阵。接着选取正、负样本像素点及其所对应的特征信息作为训练特征,采用支持向量机算法对其进行训练,得到一个分类器模型。最后将原始图像的其他所有像素点的特征信息输入到该分类器中,得到绝缘子缺陷的坐标位置。实验结果表明,提出的算法能有效地检测出输入图像中的绝缘子缺陷,验证了该改进算法的可行性及有效性。

Insulators are commonly used components in power transmission lines.Every year,accidents caused by insulator failures exceed half of power grid failures,which seriously endangers the service and operating life of the lines.Therefore,it is very important to troubleshoot insulator faults in time.It proposes an insulator defect detection algorithm based on SVM multi-feature fusion.Firstly,the algorithm first preprocesses the original image to obtain the dark channel feature layer and grayscale image,and then extracts the HSI feature layer,detail feature layer,statistical feature layer and texture feature layer on this basis,and then superimposes the above feature layers into a 3D feature matrix.Secondly,the positive and negative sample pixels and their corresponding feature information are selected as training features,and the support vector machine algorithm is used to train them to obtain a classifier model.Finally,the feature information of all other pixels of the original image is inputted into the classifier to obtain the coordinate position of the insulator defect.The experimental results show that the proposed algorithm can effectively detect the insulator defects in the input image,which verifies the feasibility and effectiveness of the improved algorithm.