Development of a Machine Vision System for Monitoring the Frothing Process

An approach to the construction of a contactless automatic control system for frothing during flotation is considered. The structure of the machine vision system is proposed. The choice of the measuring unit installation location is justified. Two approaches for constructing a contactless automatic froth control system in the flotation process based on machine vision are considered. The first is based on the traditional Watershed preprocessing and segmentation algorithm, while the second uses a convolutional neural network. The U-Net deep neural network image segmentation algorithm achieves high segmentation accuracy (IoU = 0.6947). It is resistant to noise, uneven lighting and dynamic changes in the froth.