Introduction
In the mining industry, "rock fragmentation"—the size distribution of rocks after a blast—is a critical metric that dictates downstream efficiency, from shovel digging to crusher energy consumption. While image-based analysis is the modern standard, it traditionally relies on manual photography or fixed cameras. Unmanned Aerial Vehicles (UAVs) offer a safer, automated alternative but face significant challenges in low-light environments typical of underground mines or night shifts. This paper demonstrates that UAVs equipped with artificial lighting can accurately analyze rock fragmentation in these conditions, achieving results comparable to ideal daylight.
System Architecture and Methodology
The researchers developed a comprehensive system utilizing a Parrot Bebop 2 quadrotor integrated with the Robot Operating System (ROS). The workflow involves capturing images which are then processed through MATLAB and industry-standard image analysis software to determine rock size distribution.
For indoor testing, a Vicon motion capture system provided centimeter-level positioning, while outdoor tests relied on GPS. The core of the analysis used Split-Desktop software for image segmentation to delineate rock boundaries. To validate the system, the team established a "ground truth" using physical sieve analysis on a controlled muck pile with fragments ranging from 4 mm to 19 mm. The resulting data was modeled using the Swebrec distribution function.
Experimental Results and Key Insights
A primary finding of the study is that the quality of lighting is more critical than the quantity. Uneven lighting (40 lx) often produced worse results than darker conditions due to harsh shadows creating false boundaries or bright spots causing "particle disintegration," where a single rock is seen as many.
The experiments revealed that even artificial illumination of only 14–18 lux—comparable to a parking lot at night—was sufficient to match the accuracy of daylight conditions (9500 lx), provided the light was distributed evenly.
As shown in the results, the artificial lighting configurations (Experiments 5 and 6) consistently kept the UAV's measurements within the industry-standard 30% error envelope. In outdoor "Dark" conditions (3 lx), the system failed without assistance; however, adding 14 lx of artificial light allowed the distribution curve to closely track the cloudy-day reference curve.
Conclusion and Future Directions
This research confirms that UAV-based fragmentation analysis is viable in low-light mining environments when supplemented with even artificial lighting. While the current study used fixed lighting stands, future work will focus on integrating lighting systems directly onto the UAV or secondary robotic platforms for use in inaccessible underground stopes. Additionally, scaling the system to handle massive post-blast muck piles and achieving a fully autonomous end-to-end pipeline remain key objectives for commercial implementation.
Summary updated