Advance Blasting using Drone Mapping and AI for Comprehensive Blast Performance Analysis

In mining operations, basic drone applications such as pre- and post-blast photo comparisons, blast video recordings, and sample-based fragmentation analysis are often used to assess blast performance. However, these methods lack the depth needed for comprehensive evaluations, leading to suboptimal results. This study explores the innovative integration of drone mapping technology with artificial intelligence (AI) at the Martabe Gold Mine in North Sumatra Province, Indonesia, to address these limitations. High-resolution drones and ground control points (GCPs) are utilized to create detailed orthomosaic maps, and advanced AI algorithms are applied to enhance the accuracy of blast performance insights. Field experiments reveal substantial improvements in comprehensive blast performance evaluations through this integrated approach. Enhancements include improved pre- and post-blast evaluations, AI collar detection, fragmentation size measurements, heave muckpile profiles, cast material measurements, floor prediction models, and muck pile topography volume, which is vital for ore tonnage predictions. Notably, this method achieves an average volume accuracy of 1% compared to light detection and ranging (LiDAR) technology. While LiDAR is known for its precision, the drone-based method offers a highly accurate, cost-effective, and user-friendly alternative suitable for mining applications. Despite LiDAR's superior precision, our approach provides a practical solution due to its accessibility and ease of implementation. In conclusion, the integration of advanced drone mapping and AI technologies has the potential to revolutionize blasting operations. By addressing the limitations of previous methods, this approach meets the evolving needs of the mining industry and sets new standards for accuracy and effectiveness in blast performance assessment. The combination of high-spectral drone images and AI-based software analysis demonstrated nearly 99% accuracy in ore tonnage prediction models compared to LiDAR, showcasing its promising capabilities.