城市道路土基病害核匹配追踪识别算法

发布时间：2018-03-03 04:35

  本文选题：道路工程　切入点：土基病害　出处：《中国公路学报》2017年05期 　论文类型：期刊论文

【摘要】：为准确识别道路土基病害以避免路面塌陷事故的发生,采用探地雷达对城市道路进行检测。针对城市非硬化道路和硬化道路土基病害出现的一般规律,首先通过维纳滤波器对探地雷达图像进行一维滤波,在滤波过程中,对增益函数进行倒谱域内的平滑处理,进而得到增强的雷达图像;然后,采用傅里叶变换、小波变换等信号处理方法,得到探地雷达图像在不同变换域的结果,并通过构建傅里叶核函数、多项式核函数、克罗内克核函数等不同的核函数字典,对探地雷达图像在不同变换域的结果进行匹配追踪,比较在不同核函数下的核匹配追踪序列;最后,通过比较不同核匹配追踪序列的差异,分别找出对道路土基水害和空洞病害类型敏感的核匹配追踪序列,通过该序列识别城市道路土基病害的类型。利用基于相关系数的病害度量算法比较核匹配追踪序列,判断城市道路土基病害发生的区域,并通过算法识别非硬化道路与硬化道路的地下病害。结果表明:病害类型及范围与实际情况相符,水害和空洞均能得到有效识别;通过5组城市道路探地雷达探测数据集验证了算法的有效性,识别准确率达到了99%以上;利用核匹配追踪算法处理探地雷达图像有助于城市道路土基病害识别,可减少路面塌陷事故的发生。
[Abstract]:In order to accurately identify the soil foundation diseases of the road and avoid the road collapse accident, the ground penetrating radar is used to detect the urban roads. In view of the general rules of the occurrence of the soil foundation diseases of the urban non-hardened roads and hardened roads, In the process of filtering, the gain function is smoothed in the cepstrum domain, and then the enhanced radar image is obtained. Then, Fourier transform is used to filter the GPR image. Wavelet transform and other signal processing methods are used to obtain the results of GPR images in different transform domains. Different kernel function dictionaries such as Fourier kernel function, polynomial kernel function, Croneker kernel function and so on are constructed. The results of GPR images in different transform domain are matched and tracked, and the kernel matching tracking sequences under different kernel functions are compared. Finally, the differences of different kernel matching tracking sequences are compared. The kernel matching tracking sequence, which is sensitive to the types of soil water hazards and cavity diseases, is found, and the types of urban soil foundation diseases are identified by this sequence. The correlation coefficient based disease measurement algorithm is used to compare the kernel matching tracking sequences. The area of soil foundation diseases in urban roads is judged, and the underground diseases of non-hardened roads and hardened roads are identified by the algorithm. The results show that the types and ranges of diseases are consistent with the actual situation, and both water hazards and voids can be effectively identified. The validity of the algorithm is verified by five sets of urban road ground penetrating radar data sets, and the recognition accuracy is over 99%, and the kernel matching tracking algorithm is used to deal with the ground penetrating radar images. It can reduce the occurrence of pavement collapse accident.
【作者单位】： 中国矿业大学(北京)机电与信息工程学院;中国矿业大学(北京)地球科学与测绘工程学院;
【基金】：国家自然科学基金项目(41504112) 北京市自然科学基金项目(8162035) 国家重大科学仪器设备开发专项项目(2012YQ030126) 北京市交通行业科技项目(TC1405AK9)
【分类号】：U418.5
，

本文编号：1559585