摘要
为实现岩性的快速、无损识别,基于高光谱技术与随机森林算法相结合建立了岩性的识别模型。首先,采集了3种岩石样品的高光谱,然后使用Savizkg-Golag平滑滤波、标准正态变换、多元散射校正和一阶导数4种方法对高光谱数据进行预处理,最后分别结合随机森林算法建立岩性识别模型,基于测试集准确率和F1值等方面进行综合评价,并比较偏最小二乘判别分析方法,从而选择最优模型。结果表明,经SG平滑滤波处理后建立的RF模型识别效果最佳,得到SG-RF模型测试集分类准确率为93.33%,且各类岩石对应的F1值最高。因此,高光谱技术能够实现快速、无损、准确地识别岩性,是一种很有前景的岩性识别方法。
In order to realize rapid and non-destructive lithology identification,a lithology identification model was established based on the combination of hyperspectral technology and random forest algorithm.First,the hyperspectral data of 3 kinds of rock samples were collected,and then the hyperspectral data were preprocessed by four methods:Savizkg-Golag smoothing filtering,standard nor-mal transformation,multiple scattering correction and first derivative;Finally,the lithology identification models were established by combining the random forest algorithm,and the comprehensive evaluation was carried out based on the accuracy and F1 value of the test set,and the partial least squares discriminant analysis method was compared to select the optimal model.The results show that the RF model established after SG smoothing and filtering has the best recognition effect,and the classification accuracy of SG-RF model test set is 93.33%,and the F1 values corresponding to various rocks are the highest.Therefore,hyperspectral technique is a promising method for lithology identification,which can realize rapid,nondestructive and accurate lithology identification.
作者
彭奕博
焦龙
李栋
沈瑞华
钟汉斌
熊迅宇
PENG Yibo;JIAO Long;LI Dong;SHEN Ruihua;ZHONG Hanbin;XIONG Xunyu(School of Chemistry and Chemical Engineering,Xi'an Shiyou University,Xi'an,Shaanxi 710065,China)
出处
《石油工业技术监督》
2024年第3期49-53,共5页
Technology Supervision in Petroleum Industry
基金
国家自然科学基金项目“基于新型电效团模型与多维信息融合的有机光电材料分子设计方法研究”(编号:211723003)
陕西省教育厅青年创新团队建设科研计划项目“染料敏华太阳能电池中光敏染料的计算机辅助分子设计方法研究”(编号:21JP097)。
关键词
高光谱
随机森林
偏最小二乘判别
岩性识别
hyperspectral
random forest
partial least squares discrimination
lithology identification
