基于SVM-Adaboost裂缝图像分类方法研究  

Study on Classification Method of Crack Images Based on SVM-Adaboost

作　　者：琚晓辉[1] 徐凌[2] 

机构地区：[1]交通运输部科学研究院,北京100029 [2]交通运输部公路科学研究院,北京100088

出　　处：《公路交通科技》2017年第11期38-42,共5页Journal of Highway and Transportation Research and Development

基　　金：交通运输部建设科技项目(2014318J21060)

摘　　要：不同的裂缝类型关系到不同养护策略。SVM在解决小样本、非线性、高维度问题时具有较大优势,通过采用不同的SVM分类方法和核函数对常用的数据集中的样本进行分类结果对比,选取了RBF核函数和One-against-All的分类方法。但分类结果仍然满足不了路面养护要求。由于Adaboost选择不同的样本进行训练,改变了训练样本的数据分布。每次迭代都会计算得到一个分类效果最佳的弱分类器及其所在总体分类器中的权重。随着迭代次数的增加,最终由弱分类器迭代生成的强分类器的分类误差最小。提出了SVM-Adaboost分类器动态的对SVM参数进行优化。试验结果表明,应用基于SVM-Adaboost的裂缝分类算法对指定样本进行测试,横向裂缝准确率87.48%,纵向裂缝准确率95.37%,网状裂缝准确率94.9%,块状裂缝准确率89.7%。该方法可以提高组合分类器整体的分类精度。Different classes of cracks are closely related to different maintenance strategies. As SVM has great advantages in solving small samples, nonlinear and high-dimensional problems, different core functions of RBF and the classification methods of One-against-All in SVM are selected and adopted to compare the classification results of the commonly used samples in data set, hut the classification results still cannot meet the requirements of pavement maintenance. As the Adaboost algorithm selects different samples for training, the data distribution of training samples is changed. The weak classifier which has the best classification and its weight in the whole classifier are calculated in each iteration. With the increase of the number of iterations, the final classification error of the strong classifier generated by the weak classifier iteration is minimum. A method of SVM-Adaboost classifier is proposed to dynamic optimize the SVM parameters. The result of test on the designated samples using the classification algorithm of cracks based on SVM-Adaboost shows that the accuracy of the transverse crack is 87.48% , the accuracy of longitudinal crack is 95.37% , the accuracy of net crack is 94.9%, and the accuracy of block crack is 89.7%. The proposed method carl improve the whole classification accuracy of the combined classifier.

关 键 词：道路工程 养护管理 裂缝分类 图像处理 SVM ADABOOST 

分 类 号：U418.4[交通运输工程—道路与铁道工程]