改进BTSVM算法及其在桥梁裂缝检测中的应用  

Improved BTSVM algorithm and its application in bridge crack detection

作　　者：李浩 罗伟[2] LI Hao;LUO Wei(Hunan Survey and Design Institute Co.,Ltd.,Changsha 410014,China;College of Intelligent Manufacturing,Hunan Railway Professional Technology College,Zhuzhou 412001,China)

机构地区：[1]湖南省勘测设计院有限公司,湖南长沙410014 [2]湖南铁道职业技术学院智能制造学院,湖南株洲412001

出　　处：《现代电子技术》2024年第22期119-124,共6页Modern Electronics Technique

基　　金：湖南铁道职业技术学院机电一体化科研创新团队建设资助项目(KYTD202103)。

摘　　要：针对一般的二叉树支持向量机(BTSVM)分类器收敛速度慢以及完全二叉树和偏二叉树支持向量机分类器错分样本易积累这两个问题,提出一种改进的BTSVM智能识别方法。通过构造一个非平衡二叉树SVM,减少由于前期分类错误造成的误差积累。利用飞狼优化算法对BTSVM算法两个核心参数进行寻优处理,并利用参数优化后的BTSVM对桥梁裂缝进行分类诊断研究。实验结果表明,所构建的改进BTSVM算法具有较强的自主学习能力,可将易区分的类最先分离出来,拥有更高的识别准确率。In allusion to the problems of slow convergence of general binary tree support vector machine(BTSVM)classifiers and easy accumulation of misclassified samples in full binary tree and partial BTSVM classifiers,an improved intelligent recognition method for BTSVM is proposed.By constructing an unbalanced binary tree SVM,this method can reduce the error accumulation caused by earlier classification errors.The two core parameters of BTSVM algorithm are optimized by means of the flying wolf optimization algorithm,and then the bridge cracks are classified and diagnosed by means of the optimized BTSVM.The experimental results show that the constructed improved BTSVM algorithm has strong autonomous learning ability,can separate easily distinguishable classes first,and has higher recognition accuracy.

关 键 词：桥梁裂缝检测 BTSVM 完全二叉树 偏二叉树 飞狼优化算法 参数优化 PCA降维 

分 类 号：TN911.23-34[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]