Scholarlink

<p>Ore particle size is a critical indicator of ore fragmentation and a key parameter for automated mineral processing. However, traditional ore image segmentation methods suffer from low accuracy, severe under-segmentation, and weak adaptability in complex mining scenarios, while few approaches effectively balance precision and efficiency. To fill this gap, this paper proposes a novel segmentation method that combines an improved distance transform with the watershed algorithm. After image preprocessing, seed regions are optimized using a self-developed local-maximum strategy, and watershed segmentation is further refined with morphological operations. Experimental results on seven test images show that the proposed method achieves a segmentation accuracy of 0.932, an over-segmentation rate of 0.109, and an under-segmentation rate of 0.091, which are significantly better than those of traditional watershed and UNet methods.The proposed method thus exhibits superior performance and stability in practical ore particle analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An improved watershed-based algorithm for ore image segmentation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月8日

查看全文
  • Weiwei Li,
  • Wenwen Liu,
  • Yuelong Li,
  • Hang Chen

摘要

Ore particle size is a critical indicator of ore fragmentation and a key parameter for automated mineral processing. However, traditional ore image segmentation methods suffer from low accuracy, severe under-segmentation, and weak adaptability in complex mining scenarios, while few approaches effectively balance precision and efficiency. To fill this gap, this paper proposes a novel segmentation method that combines an improved distance transform with the watershed algorithm. After image preprocessing, seed regions are optimized using a self-developed local-maximum strategy, and watershed segmentation is further refined with morphological operations. Experimental results on seven test images show that the proposed method achieves a segmentation accuracy of 0.932, an over-segmentation rate of 0.109, and an under-segmentation rate of 0.091, which are significantly better than those of traditional watershed and UNet methods.The proposed method thus exhibits superior performance and stability in practical ore particle analysis.