Research on Image Segmentation and Optimization Method of Blast Pile Ore in Open-Pit Mine

The blast bulk rate is a critical indicator of blasting effectiveness in open-pit mining, and its accurate measurement primarily depends on the precision of blast pile ore image segmentation. However, traditional image segmentation methods often struggle to deliver accurate results due to the densely packed distribution of ore fragments and the low contrast at their edges in such images. To address this issue, this study proposes a segmentation approach that combines a U-Net convolutional neural network with an improved watershed algorithm to achieve high-precision segmentation of blast pile ore images. The U-Net architecture is employed to integrate both high-level semantic and low-level spatial features, facilitating effective model training for segmenting ore images. To further refine the segmentation outputs, a distance-transform-based watershed algorithm is applied, improving boundary delineation among adjacent ore blocks. The segmentation accuracy is then quantitatively evaluated using appropriate performance metrics. Experimental results demonstrate that the proposed method effectively segments blast pile ore images in open-pit mines with high accuracy. This approach provides a reliable technical basis for the automated computation of the blast bulk rate and supports the intelligent assessment of blasting performance in open-pit mining operations.