基于NCUT优化的岩心体视图像岩粒分割方法  

Particle Segmentation Algorithm for Core Stereo Microscopic Images Based on Optimizational NCUT

作　　者：吴琼 刘衍聪[1] 伊鹏[1] 刘丹[1] 战祥华 刘拓 Wu Qiong;Liu Yancong;Yi Peng;Liu Dan;Zhan Xianghua;Liu Tuo(College of Mechanical and Electrical Engineering,China University of Petroleum,Qingdao 266580)

机构地区：[1]中国石油大学(华东)机电工程学院,青岛266580

出　　处：《计算机辅助设计与图形学学报》2018年第3期485-490,共6页Journal of Computer-Aided Design & Computer Graphics

摘　　要：为了克服传统NCUT算法运算效率低、分割准确度和自适应性差等问题,基于岩心体视图像的特征,提出适用于体视图像岩粒分割的算法.首先将检测出的高光点的亮度值用该像素点邻域中亮度最小值代替,以消除图像高光噪声;然后采用2次分水岭算法分割图像,将得到的区域映射为NCUT输入节点,以块代点减少节点数量,提高分割效率;再引入边缘梯度特征和自适应的高斯核尺度因子重新设计NCUT权值矩阵,提高分割准确度和自适应性;最后加入限制条件约束K-means初始聚类中心选择,提高NCUT聚类稳定性.实验结果表明,该算法降低了对初始参数的依赖,对岩粒有更好的分割效果,且运算时间明显缩短.In order to overcome the problems of traditional NCUT such as low efficiency,poor resolution and poor adaptability,this paper proposed an algorithm for particle segmentation based on the characteristics of core stereo microscopic images.In this paper,the detected high-light pixels were replaced by the minimum value of the luminance in the neighborhood of the pixel to eliminate the image highlight;then the second watershed algorithm was used to segment image,and the resulting regions were mapped to the NCUT input nodes,so the number of nodes were reduced by the block point,and the segmentation efficiency was improved;and also the NCUT weight matrix was redesigned by introducing edge gradient feature and using adaptive Gaussian kernel scale factors to improve segmentation accuracy and adaptability;finally,the constraint condition was used to determine the K-means initial clustering centers to improve clustering stability.All above achieved the optimization of particle segmentation algorithm.The results showed that the optimized NCUT reduced the dependence on the initial parameters,had better segmentation effect,and shortened the computation time obviously.

关 键 词：岩心体视图像 岩粒分割 2次分水岭 NCUT 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]