Abstract:

Over the years, quantitative indicators in the upstream (exploration) to downstream (mining, beneficiation, and smelting) phases of the industrial chain have primarily relied on valuable elements. However, due to the diversity of minerals hosting these elements (where a single element can be present in multiple mineral phases), relying solely on elemental quantification often leads to significant discrepancies between resource reserve assessments in the early stages and resource recovery evaluations later on. Evidently, as the beneficiation process aims to recover target minerals containing recoverable elements, it is more reasonable to use the content of target minerals, rather than target elements, as the quantitative indicator. Consequently, for recoverable target minerals, the application of artificial intelligence image recognition technology for rapid and accurate quantification merits further investigation. The Huangtun copper-gold deposit in Anhui, a hydrothermal deposit, has copper (Cu) as one of its primary recoverable resources, with chalcopyrite being the primary mineral phase hosting Cu. Through the use of deep learning methods for microscopic image analysis, chalcopyrite was identified and quantified in 114 samples from 13 boreholes across five exploration lines in the mining area. The results demonstrate that AI image recognition technology based on deep learning can accurately identify and quantify chalcopyrite, providing a more comprehensive understanding of the spatial variation and distribution of copper resources compared to elemental quantification. This has significant implications for guiding exploration, mining, beneficiation, and separation of deep copper deposits or related minerals (such as gold), ultimately enhancing the comprehensive utilization efficiency of resources.

Key words: chalcopyrite, deep learning, mineral quantification, Huangtun copper-gold deposit