CAJ | 학술논문

石英和长石的识别对储集层研究具有重要意义。传统的矿物成分分析主要依靠人机交互式识别，工作量大且效率低，针对上述问题，提出一种利用岩石颗粒在正交偏光镜下的纹理特征进行识别的方法。首先用Sobel算子提取样本图像的梯度信息，计算每个样本梯度图像的灰度共生矩阵的能量和相关性，利用能量和相关性为目标特征参数组建石英、长石特征参数样本库。应用人工神经网络(ANN)分类方法进行训练，基于训练的结果，计算待识别颗粒的特征参数并分类。最后利用偏光序列图进行决策，得出最终识别结果。实验结果表明，此识别方法对石英和长石有较好的识别效果，识别率达85%。
석영화장석적식별대저집층연구구유중요의의。전통적광물성분분석주요의고인궤교호식식별，공작량대차효솔저，침대상술문제，제출일충이용암석과립재정교편광경하적문리특정진행식별적방법。수선용Sobel산자제취양본도상적제도신식，계산매개양본제도도상적회도공생구진적능량화상관성，이용능량화상관성위목표특정삼수조건석영、장석특정삼수양본고。응용인공신경망락(ANN)분류방법진행훈련，기우훈련적결과，계산대식별과립적특정삼수병분류。최후이용편광서렬도진행결책，득출최종식별결과。실험결과표명，차식별방법대석영화장석유교호적식별효과，식별솔체85%。
Recognition of quartz and feldspar is meaningful for reservoir research. Traditional mineral composition analysis mainly relies on the man-machine interactive identification, which brings enormous work as well as low efficiency. Considering above problems, an effective dividing method is proposed in this paper based on rock particles' texture characteristics under orthogonal polarizer. The gradient information of sample images is firstly extracted using Sobel operator. And Gray Level Co-occurrence Matrix(GLCM) energy and relevance of each gradient image are calculated as the characteristic parameters sample library. Then apply Artificial Neural Network(ANN) classification methods to training. Based on the training data,the particles is identified according to the characteristic parameters. At last, polarized sequence diagram is used to decide the final recognition result. The experimental results indicate that the method of recognizing quartz and feldspar achieves good effect.