Abstract
The surface feature of flotation froth is an indicator of the flotation process state. Image-based methods have long been considered as an indirect detector to access flotation working conditions. However, large and small bubbles stick together, resulting in shadows, occlusions and defocus problems in the flotation froth images acquired in the field. These problems lead to resistance to the accurate extraction of morphological features. This paper attempts to analyze the morphology of froth images from the perspective of bubble distribution. The proposed framework generates density maps measuring the sparsity of bubbles, which also serves as a meter to imply the morphology of froth images. In order to improve the quality of the regressed density map, label normalization was proposed to calibrate ground truth during the dataset production phase; the deconvolution module was introduced to the network to gain smoother bubbles boundaries; the loss selective drop mechanism was used to mitigate the negative impact of annotation deviation during the model training. The effectiveness of each module in the framework was verified by a series of ablation experiments on the coal flotation froth dataset. The experimental data shows that the error in measuring bubble mean size is 1.7%, the error in measuring ratio of the area between large and small bubbles is 0.8%, and the accuracy of measuring the spatial distribution of large and small bubbles is satisfactory.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.