自适应聚类Hough变换及地震断层检测  

Adaptive clustering Hough transform and fault detection

作　　者：陈雷[1] 肖创柏[1] 禹晶[1] 张亚红[1] 王真理[2] 

机构地区：[1]北京工业大学计算机学院,北京100124 [2]中国科学院地质与地球物理研究所,北京100029

出　　处：《高技术通讯》2017年第3期193-202,共10页Chinese High Technology Letters

基　　金：国家自然科学基金(61501008);北京市自然科学基金(4162007)资助项目

摘　　要：针对现有基于Hough变换的地震断层检测方法只能检测单个断层,不能准确检测多个断层的不足,提出了一种基于自适应聚类Hough变换的地震断层检测方法。该方法首先对预处理后的地震相干图像进行边缘检测并对边缘图像进行Hough变换以检测出边缘图像中的线段,然后根据倾斜角和位置信息对线段进行自适应聚类以获得更完整的线段,最后根据初始地震图像对完整线段中的各点进行调整以获得准确、平滑的断层。为验证该方法的有效性,在实际地震图像上进行了对比实验。实验结果表明,该方法可正确检测地震图像中的多个断层,正确率达到90%以上,与现有方法相比,峰值信噪比提高了约10%。In consideration of the problem that the traditional Hough transform based methods for seismic fault detection can only detect single fault while multiple faults can not be detected accurately, a new fault detection method based on adaptive clustering Hough transform was proposed. The method is described below： Firstly, edge detection is performed on the pre-processed seismic coherent image and the Hough transform of the edge image is performed to detect line segments. Then, the line segments are clustered adaptively according to their inclination angle and position to obtain a more complete line segment. Finally, each point on the complete line segment is optimized according to the original seismic image to make the fault more accurate and smoother. To berify the effectiveness of the proposed method, the experiments were done on real seismic images. The experimental results indicated that multiple faults in seismic images were detected correctly by the proposed method. The correct rate was above 90%, and the peak signal noise ratio was improved by 10% compared with current algorithms.

关 键 词：HOUGH变换 地震图像 线段 断层检测 自适应聚类 

分 类 号：P631.4[天文地球—地质矿产勘探] TP391.41[天文地球—地质学]