.png){kind=icon}
.png){kind=icon}
Return
Segment anything model-based method for precise froth size determination in flotation process
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ces.2025.121657.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Automated control of the flotation process relies on indicative information from froth images, such as texture, flow rate, shape, and froth size distribution. Conventional image segmentation methods often produce errors and lack robustness when handling variations in mineral color, resolution, and contrast. Deep learning approaches, while effective, demand extensive labeled data and high training costs. This paper introduces a high-precision froth image segmentation algorithm utilizing the Segment Anything Model (SAM) visual foundation model for flotation froth image analysis. Evaluated on three diverse mineral image datasets against manual segmentation, the method demonstrates over 90% segmentation accuracy without requiring training or fine-tuning of the SAM, significantly reducing the need for labor-intensive dataset labeling and lowering application costs. The proposed approach offers a more efficient and cost-effective solution for froth image segmentation in the mining industry, enhancing automation and process optimization.
Keywords:
Flotation froth
Image segmentation
Deep learning
Segment anything model
Journal
IF:
4.3
Papers:
2.2W
Citations:
5.5W