基于支持向量机的道路地下空洞量化识别方法

Quantitative recognition of underground cavities in roadbed detection based on support vector machine

如何实现地下空洞的量化识别是探地雷达技术在城市道路地下空洞探测应用中的难题。针对城市道路地下空洞识别具有的小样本、非线性及高维模式识别特征,本文提出了一种基于支持向量机(SVM)的地下空洞量化识别方法。首先进行地下空洞正演模拟,对地下空洞雷达模拟数据进行预处理并进行数据降维,然后对降维后的数据进行时间域特征提取,利用SVM算法进行分类训练并构建空洞识别模型,最后应用该模型实现对地下空洞测试数据的量化识别。应用该方法分别对城市道路地下空洞正演模拟数据和物理模型数据进行实验验证,结果表明:本方法可以实现城市道路地下空洞的量化识别,其中正演模拟数据空洞识别精确度达到94%,空洞大小误差为±0.1m;物理模型数据空洞识别精确度达到82%,空洞大小误差为±0.2m。本方法可应用在城市道路地下空洞的量化识别中,为道路安全提供技术支撑。

It is a difficult problem for quantitative identification of the underground cavities by using ground penetrating radar technique in urban road. Aiming at the small sample, nonlinear and highdimensional pattern recognition features for urban road underground cavity identification, an underground cavity quantitative identification method is proposed based on support vector machine ( SVM). Firstly, the forward simulation of underground cavity is performed and the forward simulation data is preprocessed and dimension-reduced , and time domain feature extraction is carried out on the reduced dimension data. Then the SVM algorithm is used for classification training and the cavity identification model is constructed, which is used for the application. The verification experiments are implemented for forward model data and physical model data by using this proposed method, and the results show that this method can realize the quantitative identification of underground cavities in urban roads, and the accuracy of cavities identification for forward simulation data reaches 94% wiith error at ±0. 1 m, and the accuracy for physical model data reaches 82% with error at ±0. 2m. This method can be applied for the quantitative identification of road underground cavities, providing technical support for road safety.