Abstract

This study presents a multimodal data fusion system to identify and impact rocks in mining comminution tasks, specifically during the crushing stage. The system integrates information from various sensory modalities to enhance data accuracy, even under challenging environmental conditions such as dust and lighting variations. For the strategy selected in this study, 15 rock characteristics are extracted at neighborhood radii of 5 mm, 10 mm, 15 mm, 20 mm, and 25 mm to determine the suitable impact points. Through processes like the Ball-Pivoting Algorithm (BPA) and Poisson Surface Reconstruction techniques, the study achieves a detailed reconstruction of filtered points based on the selected characteristics. Unlike related studies focused on controlled conditions or limited analysis of specific rock shapes, this study examines all rock faces, ensuring the more accurate identification of impact points under adverse conditions. Results show that rock faces with the largest support areas are most suitable for receiving impacts, enhancing the efficiency and stability of the crushing process. This approach addresses the limitations of manual operations and provides a pathway for reducing operational costs and energy consumption. Furthermore, it establishes a robust foundation for future research to develop fully autonomous systems capable of maintaining reliable performance in extreme mining environments.