| 于立红,张善良,王国君.基于多源地学数据的找矿预测应用研究[J].矿产勘查,2023,14(8):1432-1439 |
| 基于多源地学数据的找矿预测应用研究 |
| Application of prospecting prediction based on multi-source geoscience data |
| 投稿时间：2022-03-02 修订日期：2023-06-10 |
| DOI：10.20008/j.kckc.202308010 |
| 中文关键词: 卷积神经网络模型 生成式对抗网络模型 化探数据 航磁数据 断裂构造数据 找矿预测 |
| 英文关键词: convolution neural network model generative countermeasure network model geochemical exploration data aeromagnetic data fault data prospecting prediction |
| 基金项目:本文受新疆维吾尔自治区地质勘查基金项目“新疆东天山吉木萨尔林场一带1∶5万K45E002019、K45E002020、K45E002021、K45E002022四幅区域地质矿产调查”（T14-1-LQ08）资助。 |
|
| 摘要点击次数: 1007 |
| 全文下载次数: 505 |
| 中文摘要: |
| 为提高地质找矿精度，本文结合研究区化探、航磁、地质以及断裂构造数据，运用卷积神经网络（CNN）模型，对基于多种数据的铜矿床预测结果进行了对比分析。结果表明：在运用卷积神经网络（CNN）模型进行找矿预测前，需要利用克里格插值法对化探和航磁数据进行预处理；同时传统地质数据断裂构造解释不详细的问题，需利用生成式对抗网络（GAN）模型对遥感影像数据进行断裂构造数据的解译，从而获得多种地质数据。基于化探、航磁、地质以及断裂构造数据预测得到的铜矿床面积分别占研究区的27.3%、12.1%和19.7%；由于地质数据仅包括研究区的主干断裂，导致在预测过程中某些重点找矿区未被圈定，而采用GAN模型对断裂构造数据进行训练预测后，未被圈定的区域被重新圈定，将断裂构造数据+化探航磁数据相结合应用到铜矿床预测模型中，具有更好的预测效果和可信度。 |
| 英文摘要: |
| In order to improve the accuracy of geological prospecting， the convolutional neural network （CNN） model is used to compare and analyze the prediction results of copper deposits based on various data combined with geochemical exploration， aeromagnetic， geological and fault data in the study area. The results show that before using convolutional neural network （CNN） model for prospecting prediction， it is necessary to use Kriging interpolation method to preprocess geochemical exploration and aeromagnetic data. At the same time， the fault interpretation of traditional geological data is not detailed， so it is necessary to use generative adversarial network （GAN） model to interpret the fault data of remote sensing image data， so as to obtain a variety of geological data. The predicted copper deposit area based on geochemical exploration， aeromagnetic， geological and fault data accounts for 27.3%， 12.1% and 19.7% of the study area， respectively. Due to the fact that the geological data only included the main faults in the study area， some key prospecting areas were not delineated in the prediction process. However， after the training and prediction of fault data by using GAN model， the undelineated areas were re-delineated. The combination of fault data and geochemical exploration aeromagnetic data was applied to the copper deposit prediction model， which has better prediction effect and reliability. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |