摘要
高光谱数据是一种多维遥感数据,能在可用近红外波段范围内准确采集碳酸根的振动频率数据,从而对识别碳酸盐岩岩性产生特殊效果。本研究针对传统碳酸盐岩识别方法效率低、准确度不高且受主观影响因素大等不足,使用实测得到的高光谱数据,综合对比多元线性回归模型(MLR)、主成分回归模型(PCA)、偏最小二乘回归模型(PLSR)和逐步回归模型(SR)的精度,提取最优特征,构建高光谱岩性识别模型。结果表明主成分分析模型在光谱采样间隔为2 nm时,具有较高精度(79.78%);随着光谱采样间隔的增加,模型精度均基本表现为先增加,后下降的变化趋势;当光谱采样间隔较大时,模型精度均低于70%;光谱数据采样间隔对模型精度影响较大。
Hyper-spectrum data is a multi-dimensional remote sensing data that can accurately collect the vibration frequency data of carbonate roots within the range of near-infrared bands,thereby generating special effects for identifying carbonate.Aiming at the shortcomings of traditional carbonate rock identification methods,such as low efficiency,low accuracy and large subjective factors,this study adopted four methods to construct the model by measuring the hyperspectral data of 340 carbonate rock samples collected in the field,followed by spectral preprocessing and spectral resampling.The results showed that the principal component analysis model has the highest model accuracy(79.78%),when the spectral sampling interval was 2 nm;With the increase of spectral sampling interval,the accuracy of the four models basically increased first and then decreased;When the spectral sampling interval was large,the accuracy of all models was lower than 70.00%.
作者
余秋实
邵燕林
曾齐红
魏薇
YU Qiushi;SHAO Yanlin;ZENG Qihong;WEI Wei(School of Geosciences,Yangtze University,Wuhan 430000,China;Institute of Petroleum Exploration&Development,Beijing 100083,China)
出处
《激光杂志》
CAS
北大核心
2023年第1期27-31,共5页
Laser Journal
基金
湖北省教委科研基金(No.2018289)
湖北省高等学校实验室研究项目(No.HBSY2018-28)
中石油科技项目(No.KT2021-06-14)。
关键词
碳酸盐岩
高光谱
建模
岩性
carbonate
hyper-spectral
modeling
lithology
