基于改进级联Gentle Adaboost分类器的支柱绝缘子红外图像AI识别

AI Recognition of Post Insulator Infrared Image Based on Improved Cascade Gentle Adaboost Classifier

支柱绝缘子是变电站中重要的部件,在复杂工作环境下极易出现故障,而传统人工检测难以对大量支柱绝缘子红外图片快速多目标识别。为此,提出了基于改进级联Gentel Adaboost分类器的支柱绝缘子红外图像人工智能(artificial intelligence,AI)识别方法,使用现场采集的大量红外图片,构建支柱绝缘子红外数据集,然后计算支柱绝缘子数据集样本的Haar-like特征值,并将不同特征值构建成若干个弱分类器;通过改进Gentel Adaboost算法,将弱分类器训练集成为强分类器,得到级联Gentel Adaboost分类器,实现红外图像中支柱绝缘子多目标准确识别。研究结果证明,所提方法在不同背景下对红外图像支柱绝缘子识别的准确率达到了93.9%,在正确识别定位的同时,还能保留支柱绝缘子的红外温度特征,可为支柱绝缘子智能识别和故障诊断提供有效途径。

Post insulator is an important component in substation,which is prone to failure in complex working environment.However,traditional manual detection is difficult to identify a large number of post insulator infrared images quickly.For this reason,this paper proposes an artificial intelligence(AI)recognition method of post insulator infrared image based on improved cascade Gentel Adaboost classifier.A large number of infrared images collected in the field are used to construct the post insulator infrared data set,then the Haar-like eigenvalues of the post insulator data set are calculated,and different eigenvalues are constructed into several weak classifiers.By improving the Gentel Adaboost algorithm,the weak classifier training is integrated into the strong classifier,and the cascade Gentel Adaboost classifier is obtained to realize the multi-target accurate recognition of post insulator in the infrared image.The results show that the proposed method can be adopted to recognize the post insulator in different backgrounds with 94.2%accuracy,in the meantime,the infrared temperature characteristics of post insulator can be retained while correctly identifying and locating,which provides an effective way for intelligent identification and fault diagnosis of post insulator.