基于LBP算法的SVM煤矸识别  

SVM coal-gangue recognition based on LBP algorithm

作　　者：李伟 刘化广[1] Li Wei;Liu Huaguang(School of Mining Engineering,Heilongjiang University of Science&Technology,Harbin 150022,China;Heilongjiang Ground Pressure&Gas Control in Deep Mining Key Lab,Heilongjiang University of Science&Technology,Harbin 150022,China)

机构地区：[1]黑龙江科技大学矿业工程学院,哈尔滨150022 [2]黑龙江科技大学黑龙江省煤矿深部开采地压控制与瓦斯治理重点实验室,哈尔滨150022

出　　处：《黑龙江科技大学学报》2023年第2期153-158,166,共7页Journal of Heilongjiang University of Science And Technology

基　　金：黑龙江省自然科学基金资助项目(ZD2021E006)。

摘　　要：为研究基于LBP算法的SVM煤矸图像识别的可行性,在实验室环境下基于Python和OpenCV2开发了煤矸图像可视化识别系统。基于采集的煤和矸石图像,通过LBP算法提取煤和矸石的特征纹理信息,建立煤和矸石特征纹理信息样本集,训练SVM模型,绘制煤矸识别散点图,实现煤和矸石的图像识别。结果表明:当用户自定义分割尺寸分别为64×64像素、128×64像素、256×128像素和256×256像素时,以多项式核和高斯核为内核的SVM模型识别效果较好,平均辨识度达90%以上;高斯核SVM模型所需惩罚系数小,且识别效果优于多项式核SVM模型,高斯核SVM模型的训练集和验证集正确率均超过93%,最高分别约为95.5%和94.4%,煤矸识别效果良好。This paper aims to study the feasibility of SVM coal gangue image recognition based on LBP algorithm.The study involves developing a visual recognition system of coal gangue image in the laboratory based on Python and OpenCV2,extracting the feature texture information of coal and gangue with LBP algorithm,based on the collected coal and gangue images;forming the sample set of coal and gangue feature texture information,drilling the SVM model,drawing the scatter plot of coal and gangue recognition;and recognizing coal image and gangue image.The results show that the SVM models with polynomial kernel and Gaussian kernel performs better recognition effect and the average recognition degree is over 90%,as happens when the user-defined segmentation size is 64×64 pixels,128×64 pixels,256×128 pixels and 256×256 pixels;the Gaussian kernel SVM model needs less penalty coefficient,and the recognition effect is better than the polynomial kernel SVM model,the accuracy of the training and validation sets of Gaussian kernel SVM model is more than 93%,contributing to a better recognition of coal gangue by the maximum of about 95.5%or 94.4%respectively.

关 键 词：煤矸识别 LBP 支持向量机 图像处理 

分 类 号：TD94[矿业工程—选矿] TP391[自动化与计算机技术—计算机应用技术]