岩石薄片正交偏光融合图像的颗粒分割方法  

Particle segmentation method based on orthogonal polarization fusion image of rock slice

作　　者：钟逸 熊淑华[1] 滕奇志[1] 吴耀坤 龚剑 Zhong Yi;Xiong Shuhua;Teng Qizhi;Wu Yaokun;Gong Jian(Institute of Image Information,School of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China;Chengdu Xitu Technology Company Limited,Chengdu 610065,China)

机构地区：[1]四川大学电子信息学院图像信息研究所,四川成都610065 [2]成都西图科技有限公司,四川成都610065

出　　处：《信息技术与网络安全》2019年第10期87-92,共6页Information Technology and Network Security

基　　金：国家自然科学基金(61372174)

摘　　要：岩石薄片中的矿物颗粒提取是自动化薄片鉴定的基础,直接影响到石油地质勘探测量的准确度。由于实际工业矿场生产中得到的岩石薄片极为复杂,采用自动化技术很难将其中的矿物颗粒完整地提取出来。针对这一问题,提出了对固定视域下的多角度正交偏光序列图进行最大值融合,利用基于Fast Minimum Barrier Distance(Fast-MBD)变换的超像素算法分割提取岩石薄片中的矿物颗粒目标,再通过最佳阈值筛选出过分割的颗粒,最后根据颜色相似性合并目标颗粒区域,实现颗粒自动分割。该方法明显改善了目前岗位上所使用方法提取效果不完善的问题,可以更好地提取岩石薄片图像中的矿物颗粒,获得较好的效果。The extraction of mineral particles from rock slice is the basis for automated slice identification,directly affecting the accuracy of petroleum geological survey measurements.Due to the complexity of the rock slice obtained in the actual industrial mine production,it is difficult to extract the mineral particles completely by using automation technology.Aiming at this problem,the maximum fusion of multi-angle orthogonal polarization sequence images in a fixed field of view is proposed.The superpixel algorithm based on Fast Minimum Barrier Distance (Fast-MBD) transform is used to segment and extract the target of mineral particles in the rock slice.Then the optimal threshold screening of the over-segmented region obtained by the superpixel algorithm is performed.Finally,the target particles are merged according to the color similarity to achieve the automatic segmentation of particles.The method obviously improves the problem that the extraction method used in the current position is not perfect,and can better extract the mineral particles in the rock slice image,further improves the accuracy of the mineral particle extraction and obtains better effects.

关 键 词：矿物颗粒分割 正交偏光图像 Fast-MBD 超像素 目标合并 最佳阈值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]