Classification of Alteration Zones Based on Drill Core Hyperspectral Data Using Semi-Supervised Adversarial Autoencoder: A Case Study in Pulang Porphyry Copper Deposit, China

With the development of hyperspectral technology, it has become possible to classify alteration zones using data. Since various altered rocks are comprehensive manifestations mineral assemblages, their spectra highly similar, which greatly increases difficulty distinguishing among them. In this study, a Semi-Supervised Adversarial Autoencoder (SSAAE) was proposed zones, drill core data collected from Pulang porphyry copper deposit. The multiscale feature extractor first integrated into encoder fully exploit and mine latent representations data, were further transformed discrete class vectors classifier. Second, decoder reconstructed original inputs with vectors. Third, we imposed categorical distribution on represented in one-hot form adversarial regularization process incorporated supervised classification network better guide training limited labeled comparison experiments synthetic dataset measured conducted quantitatively qualitatively certify effect method. results show that SSAAE outperformed six other methods for classifying zones. Moreover, displayed delineated cross-section, sensible geological point view had good spatial consistency occurrence Cu, demonstrates applicability