Flotation bubble size distribution detection based on semantic segmentation

摘要

In mineral foam flotation, characteristic information of bubble images is closely related to the flotation process condition and production indicators, among which the bubble size distribution is a key feature. Therefore, accurately obtaining the bubble size distribution is of great significance for optimizing and controlling the flotation process. Froth segmentation is the main technique to obtain bubble size information of flotation cells. In this paper, one semantic segmentation algorithm is adopted to segment flotation froth image, and a new froth segmentation algorithm based with improved U-Net is proposed. The experiment results indicate that the developed algorithm not only successfully solves the problem of over-segmentation and under-segmentation, but also improves the segmentation accuracy, which is suitable for different process conditions and lighting conditions.