论文部分内容阅读
针对传统矿物含量测定中存在人为误差、缺乏精度评价等问题,提出了基于图像分类的矿物含量测定及精度评价方法,该方法通过统计分类后图像中每种矿物的像元数量测定矿物含量,并采用混淆矩阵评价含量测定精度.根据岩石图像的光谱和纹理特征,提出了两种基本的矿物含量测定方式:1)对于纹理简单、矿物光谱区分度大的岩石图像,采用直接分类方式测定矿物含量,花岗岩手标本照片矿物分类实验表明监督分类效果优于非监督分类,且监督分类中最大似然法分类(MLC)的精度最高,其含量测定精度为94.25%;2)针对复杂纹理(如干涉色、双晶等)的岩石图像,引入了面向对象(矿物或矿物集合体)的多尺度图像分割算法,在分割基础上分类并统计每类矿物含量.白云母二长花岗岩镜下照片矿物分类实验得到其含量测定精度为94.85%.
Aiming at the problems of human error and lack of accuracy evaluation in the determination of traditional mineral content, a method for the determination of mineral content and the accuracy evaluation based on image classification is proposed. The method determines the mineral content by counting the number of pixels of each mineral in the classified image The confusion matrix was used to evaluate the accuracy of the determination of content.According to the spectral and texture features of the rock images, two basic methods for determining the mineral content were proposed: 1) For the simple texture and large discriminating degree of mineral spectrum, the direct classification method was used to determine the mineral content , And the mineral classification experiments of the photos of hand samples of granite show that the supervised classification is superior to unsupervised classification, and the maximum likelihood classification (MLC) of the supervised classification has the highest accuracy with a content determination accuracy of 94.25%. 2) For complex textures such as interference Color, twin, etc.), an object-oriented (mineral or mineral assemblage) multi-scale image segmentation algorithm is introduced to classify and count each type of mineral content based on the segmentation. Experiments to determine the content of its accuracy of 94.85%.