Three-dimensional quantitative mineral prediction from convolutional neural network model in developing intelligent cleaning technology

The aim of this study is to explore a three-dimensional (3D) quantitative mineral prediction method to address the issues of low accuracy and efficiency in mineral resource exploration. The experiment constructs a 3D mineral image prediction model based on intelligent clean technology, incorporating an attention convolutional neural network (CNN). This model first introduces the ShuffleNet V2 network, a lightweight and efficient CNN renowned for handling complex geological resource image data. Additionally, the model incorporates graph attention modules and channel attention modules to enhance the network's focus on crucial channel information, enabling better extraction of spatiotemporal features from 3D mineral resource samples. The results show that, compared to CNN algorithms, the accuracy of the proposed model in 3D mineral identification reaches 95.25%, a minimum accuracy improvement of 1.41%. Moreover, under induced fault geological types, this model achieves a Mean Intersection over Union (mIoU) value of 94.55%. The constructed model demonstrates high accuracy and precision in prediction performance and sustainability, providing strong support for the sustainable development and strategic direction of mineral resource exploration.

Automated summary

The aim of this study is to explore a three-dimensional quantitative mineral prediction method that addresses the low accuracy and efficiency issues in mineral resource exploration. The experiment constructs a 3D mineral image prediction model incorporating an attention convolutional neural network (CNN). The results show that the proposed model achieves higher accuracy and precision in 3D mineral identification compared to CNN algorithms, and it demonstrates strong support for the sustainable development and strategic direction of mineral resource exploration.