煤岩识别算法及“自学习”模型研究

Research on coal rock recognition algorithm and “self-learning” model

针对煤矿井下综采工作面煤岩识别方法普遍存在效果欠佳、稳定性差、应用范围小等问题,基于煤和岩石基本特征的区别,从图像边缘和灰度阈值等视觉差异,借助聚类基本理论,处理煤岩图像边界,分析煤岩灰度共生矩阵包含的纹理特征信息;构造均值纹理导向度和方差纹理导向度;研究煤岩界面灰度共生均值的聚类煤岩识别算法;多尺度分解变换的煤岩纹理特征抽取方法;提出2种分块区域煤岩分界的图像识别模型:图像灰度"相似性"度量估计模型、层次聚类识别模型;并构造出一种煤岩混合模型融合识别方法及流程;为工作面煤岩精确识别、减少采煤机截割岩引起的故障和安全问题提供参考。

Aiming at the problems of poor effect, poor stability and small application range of coal and rock recognition methods in fully mechanized coal mining face, based on the difference between the basic characteristics of coal and rock, from the visual differences of image edge and gray threshold, with the help of clustering theory, image processing is used to analyze the boundary of coal and rock image, and the textural feature information contained in coal and rock gray level co-occurrence matrix is analyzed in this paper, two image recognition models of coal rock boundary are proposed: the"similarity"measurement estimation model of image gray scale and the Gaussian mixture clustering recognition model;and a coal rock mixed model is constructed. It provides a reference for the accurate identification of coal and rock in working face and the reduction of faults and safety problems caused by shearer cutting rock.