Feature extraction of static foam images based on U-Net and Hough transform

In flotation-based mineral processing, froth images contain important visual information that reflects the operating condition of the process and provides a basis for process monitoring and control. However, conventional machine-vision methods still encounter difficulties in extracting static froth features under conditions of bubble adhesion, overlap, weak boundaries, and uneven illumination. To address these problems, a flotation froth image analysis framework is proposed by combining a boundary-aware U-Net segmentation model with a segmentation-guided adaptive Hough transform. In the first stage, the U-Net architecture is enhanced with explicit boundary-oriented supervision to improve the separation of adherent and overlapping bubbles. In the second stage, the geometric priors derived from the segmentation results are used to constrain the Hough-based detection process, thereby improving the reliability of bubble counting and geometric feature extraction. Experimental results show that the proposed method achieves a Dice similarity coefficient of 96.39% and an intersection over union of 93.04%, while maintaining favorable inference efficiency. Compared with representative segmentation models, the proposed framework provides a better balance between accuracy and speed. The study demonstrates that the integration of deep boundary-aware segmentation and adaptive geometric detection is effective for static feature extraction of flotation froth images and has potential for intelligent monitoring of flotation processes.