Application Of Froth Images Classification And Clustering Based On Visual Features In Flotation Cell Performance

Flotation is the most frequently approach for beneficiation of metallic ores in mineral processing plants. Continuous control of flotation circuits is necessary to achieve optimum metallurgical performance. Previous research has established that there is a meaningful correlation between the froth visual features and process conditions and performance. The main objective of the current study is to develop algorithms for extraction of visual (bubble size, froth velocity and froth colour) and textural (energy, entropy and correlation) features from the froth images as well as classification of the images based on the captured properties. For this purpose, flotation tests were conducted in a batch cell under various process conditions and the metallurgical parameters (copper recovery and concentrate grade) along with the image variables were measured. Decision tree and fuzzy C-means algorithms were used for classification and clustering of the froth images. It was found that the developed machine vision system is capable of more accurately classifying the froth images than a manual operatory system. The results indicate that the developed algorithms are capable of accurately classifying the froth images with respect to the visual as well as the metallurgical parameters, which is of central importance for development of a machine vision based control system.