基于卷积神经网络的智能找矿预测方法：以甘肃龙首山地区铜矿为例

智能找矿预测是数字地质科学的前沿领域。本文基于一种二维卷积神经网络的智能找矿预测方法，以25种元素的水系沉积物数据和航磁数据为找矿预测数据，将已知的矿点作为监督样本，利用步长平移数据增强方法获取了训练数据集，对卷积神经网络进行训练后，将其应用于未知区域的找矿预测。应用该方法对甘肃省龙首山西段高台县臭泥墩—西小口子地区进行了铜矿智能找矿预测，根据已知的3个铜矿点，获取了22 934个训练数据，经过200轮训练之后，预测精度能够达到98.1%，最终圈定了5个预测区，5个预测区均具有良好的铜矿找矿远景。

Intelligent Prospect Prediction Method Based on Convolutional Neural Network:A Case Study of Copper Deposits in Longshoushan Area,Gansu Province

Intelligent prospect prediction method is the leading edge of digital geoscience. In this paper, an intelligent prospect prediction method based on two-dimensional convolutional neural network is used. On the basis of 25 elements and aeromagnetic data on geochemical survey of stream sediments and taking known ore occurrences as monitoring samples, the training data set is obtained by step shift data enhancement. After training the convolutional neural network, it is applied to prospect prediction of unknown areas. The intelligent prospecting of copper deposits in the area of Chounidun-Xixiaokouzi, Gaotai County, west Longshoushan, Gansu Province is predicted. From 3 known copper occurrences, 22 934 training data are obtained. After 200 rounds of training, the prediction accuracy reaches 98.1%, and 5 prediction areas are delineated. In consideration with the previous research results and field work, the delineated areas have good prospect for copper mineralization.