###
DOI:
工程科学与技术:2012,44(6):203-210
←前一篇   |   后一篇→
本文二维码信息
码上扫一扫!
基于数字图像处理和SVM的岩体裂隙迹线自动检测
(1.四川大学 制造科学与工程学院;2.西北核技术研究所)
Automated Detection of Rock Discontinuity Trace Based on Digital Image Processing and SVM
(1.School of Manufacturing Sci. and Eng.,Sichuan Univ.;2.Northwest Inst. of Nuclear Technol.)
摘要
图/表
参考文献
相似文献
本文已被:浏览 2137次   下载 4
投稿时间:2012-04-18    修订日期:2012-07-09
中文摘要: 提出了一种基于线特征检测、SVM裂隙识别和分割迹线自动连接的岩体裂隙迹线自动检测新方法,该方法计算图像光强度函数2阶导数最大值方向上每个像素点的1阶导数,将1阶导数过零点标识为线特征点,按边缘相似性原则连接成线分段。用线分段的光度参数和几何学参数作为描述裂隙迹线的特征参数,采用盒约束的软间隔优化方法实现裂隙迹线的识别分类。实验结果显示,所提出的方法,可从岩体暴露面图像中自动检测识别裂隙迹线,自动生成的迹线图与地质工作人员手工绘制的迹线图基本相符,表明了本文方法的有效性。
Abstract:A new discontinuity trace automated detection methodology was presented based on automated line feature detection,SVM fracture recognition, and segmentation tracing linking. The line feature point was the zero crossings of the light intensity function’s first derivatives, which were calculated at each pixel in the direction where the image light intensity function’s second derivative was maximum. All line feature points were linked into line segmentations according to the principle of edge likelihood. These line segmentations were thus characterized by calculating a series of photometric and geometrical parameters. Then, the fracture traces were classified by the soft margin optimization with box constraints. The experimental results showed that the fracture trace in the rock mass exposure image can be automated detected and recognized by this new methodology. The discontinuity trace map constructed by the new methodology is mainly consistent with the map drawn manually. The results demonstrated the effectiveness of the proposed methodology.
文章编号:201200282     中图分类号:    文献标志码:
基金项目:科研预研项目资助项目(KJ2011020;KY201002B)
作者简介:
引用文本:
郭立钱,廖俊必,钟方平,陈剑杰,黄昊,余翔,董开营.基于数字图像处理和SVM的岩体裂隙迹线自动检测[J].工程科学与技术,2012,44(6):203-210.
Guo Liqian,Liao Junbi,Zhong Fangping,Chen Jianjie,Huang Hao,Yu Xiang,Dong Kaiying.Automated Detection of Rock Discontinuity Trace Based on Digital Image Processing and SVM[J].Advanced Engineering Sciences,2012,44(6):203-210.