考虑几何特征的岩体裂隙迹线多尺度图像识别方法  

Multi-scale image recognition method of rock mass fracture traces considering geometric features

作　　者：吴顺川[1,2] 孙贝贝 余一松 王勇兵[4] WU Shunchuan;SUN Beibei;YU Yisong;WANG Yongbing(Key Laboratory of Ministry of Education for Efficient Mining and Safety of Metal Mine,University of Science and Technology Beijing,Beijing 100083,China;Faculty of Land Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;Changsha Engineering&Research Institute Ltd.of Nonferrous Metallurgy,Changsha 410019,China;Yunnan Chihong Zn&Ge Co.Ltd.,Qujing 654200,Yunnan China)

机构地区：[1]北京科技大学金属矿山高效开采与安全教育部重点实验室,北京100083 [2]昆明理工大学国土资源工程学院,云南昆明650093 [3]长沙有色冶金设计研究院有限公司,湖南长沙410019 [4]云南驰宏锌锗股份有限公司,云南曲靖654200

出　　处：《华中科技大学学报（自然科学版）》2024年第9期101-109,117,共10页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(51934003);云南省重大科技项目(202202AG050014)。

摘　　要：针对以往研究中岩体裂隙图像识别方法的识别结果存在大量伪裂隙问题,提出了一种考虑几何特征的岩体裂隙迹线多尺度图像识别新方法.根据图像中不同像素点黑塞矩阵特征值大小的差异构建滤波响应函数,该函数在对线性裂隙进行增强的同时亦可抑制图像噪声;采用限制对比度的直方图均衡化算法解决了函数响应值不足的问题;基于连通域面积和连通域最小外接矩形长宽比两个参数准确获取了裂隙形态;利用形态学细化、毛刺去除和分段线连接的组合处理方式识别得到裂隙迹线.工程实测结果表明:所提方法裂隙迹线识别结果与手工描绘的参考裂隙极为相符,精度和敏感度的调和平均值D_(ice)达到88%以上.相较于改进Canny方法、暗区曲线结构增强(DRCSE)方法以及混合全局和局部阈值(HGLT)方法,本文方法表现出了更强的稳健性和鲁棒性.In address the problem of a large number of pseudo-fractures in the recognition results of the image recognition methods for rock mass fractures in previous studies,a multi-scale image recognition method of rock mass fracture traces considering the geometric features was proposed.A filter response function was constructed based on the difference in the magnitude of the eigenvalues of the Hessian matrix of different pixels,which can enhance the linear fractures while suppressing the image noise.The problem of insufficient response value of the function was solved by using the contrast limited adaptive histogram equalization algorithm.The fracture geometry was accurately obtained based on the connected domain area and length-width ratio of the minimum external rectangle of the connected domain.Fracture traces were identified by morphological thinning,burr removal and segment line linking algorithms.The engineering results show that the fracture traces identified by the proposed method are the most consistent with the reference fractures drawn by hand,and the D_(ice) values are more than 88%.Compared to the improved Canny method,dark region curvilinear structure enhancement(DRCSE) method and hybrid global and local threshold(HGLT)method,the proposed method shows better performance.

关 键 词：岩体 裂隙迹线 图像识别 HESSIAN矩阵 图像处理 

分 类 号：TU45[建筑科学—岩土工程]