Multivariate Geochemical Anomaly Identification Method Based on Spatial Constraint Multiple Autoencoders

Abstract

The invention discloses a multivariate geochemical anomaly identification method based on multiple self-encoders with space constraints, comprising the following steps: data clustering; space domain division; self-supervised learning; abnormal value calculation; and abnormal map generation. The present invention combines multivariate data clustering, spatial filtering and self-encoding neural network, taking into account complex relationships and spatial distribution characteristics among various geochemical elements, thereby effectively improving the accuracy of geochemical anomaly identification and providing a solution for utilizing multivariate geochemical elements under complex geological conditions. The anomaly identification of geochemical prospecting data provides a more practical and reliable scientific method.

Description

technical field [0001] The invention relates to the field of geochemical anomaly identification and the field of artificial intelligence application, in particular to a method for multivariate geochemical anomaly identification based on spatially constrained multi-autoencoders. Background technique [0002] Geochemical anomaly identification is one of the important tasks in mineral exploration. Geologists combine geochemical anomalies with other geological information to make comprehensive judgments to find potential ore fields. Abnormal identification methods based on frequency domain are widely used, such as box plot method, mean method, multivariate data analysis method, etc. But these methods lack to consider an important characteristic of geochemical data---the spatiality of geochemical field. Fractal / multifractal, kriging method, singular value method, spatial factor analysis method, trend surface method, etc., consider the correlation of spatial neighbor samples, and...

AI Technical Summary

Problems solved by technology

However, because the geochemical anomaly identification method based on a single neural network model lacks consideration of regional geochemical background differences and cannot distinguish spatial differences according to local conditions, it is difficult to train model parameters, and the identification of potential mineral resource areas containing weak anomalies is accurate. low rate

In short, there is currently a lack of geochemical anomaly identification methods that comprehensively consider spatial differences in geochemical background, non-normal distribution of data, and nonlinear relationships among multiple chemical elements

Images