岩石颗粒图像中的斜长石识别  

Plagioclase Recognition of Complex Texture in Rock Particles

作　　者：周磊 滕奇志[1] 何海波 

机构地区：[1]四川大学电子信息学院图像信息研究所,成都610065 [2]成都西图科技有限公司,成都610024

出　　处：《现代计算机》2018年第2期19-25,共7页Modern Computer

基　　金：四川大学研究生课程建设项目(No.2016KCJS5113)

摘　　要：针对具有复杂纹理岩石颗粒目标图像中无法准确提取直线纹理特征的问题,提出一种基于LSD候选直线段分段拟合的方法。该算法利用LSD局部直线检测算法,将候选断裂直线段进行多段相似拟合。首先根据图像梯度信息,采用动态阈值的区域生长策略获得具有相似属性的连通区间,再对各个连通区间做外接矩形得到候选直线段,并对提取后的多条候选直线段进行分段相似拟合,最后根据拟合后的直线段的二维特征,进行斜长石纹理特征的识别。实验结果表明该算法在复杂纹理岩石颗粒目标上的直线检测均优于传统的霍夫变换和LSD局部直线检测,同时对斜长石纹理特征的识别具有较高的准确率。In terms of the problem of extracting linear texture features from rock particles with complex textures is difficult to extract, proposes a meth- od of segment fitting of candidate straight line segments based on LSD. This algorithm is based on the LSD local line detection algorithm, fits the cracked candidate line segments. Firstly, according to the image gradient information, uses a region growing strategy with dynamic threshold to obtain the connected region with similar attributes, then uses the outer connected rectangle to obtain the candidate straight line segment, and the segmented candidate lines are fitted similarly, finally, according to the two-dimensional features of the line segments, identifying plagioclase texture features. The experimental results show that the texture feature recognition of plagioclase is superior to the traditional Hough transformation and LSD local line detection algorithm in complex texture the rock particles on the target, at the same time, the recognition of plagioclase texture features a higher accuracy.

关 键 词：斜长石识别 矿物显微图像 直线检测 LSD局部直线检测 相似直线段拟合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]