基于改进HOG特征提取与SVM分类器输电线路金具识别

Transmission line hardware identification based on improved HOG feature extraction and SVM classifier

目的:针对识别输电线路金具的检测方法存在精度低、分类结果较差的问题,提出解决方案。方法:提出一种基于支持向量机(SVM)分类与改进HOG梯度方向直方图特征提取相结合的输电线路金具识别算法。对图像进行去噪预处理,提取特征信息,输入到SVM进行识别分类。将绝缘子规定为正样本,耐张线夹规定为负样本,选取500个样本进行试验。结果:准确率由未改进前的81%提升到96%。结论:所提出的算法可行、有效,为机器学习在输电线路金具识别中的应用提供一定的参考。

Objective:The current detection methods for identifying transmission line fittings have the problems of low accuracy and poor classification results.Methods:To solve this problem,this paper proposes a transmission line hardware recognition algorithm based on support vector machine(SVM)support vector machine classification and improves HOG gradient direction histogram feature extraction.The image as pretreatment was denoised.The feature information was extracted,which was input to the SVM for recognition and classification.The insulator is specified as a positive sample and the tension clamp is specified as a negative sample,500 samples are selected for the experiment.Results:The accuracy is increased from 81%before improvement to 96%.Conclusion:The feasibility and effectiveness of the experimental method are verified.The method provides a certain reference for the application of machine learning in the identification for transmission line fittings.