Production conditions classification for froth flotation based on froth image processing

The visual parameters of the froth surface are significant to the flotation production statuses, but they are difficult to be measured with tradition image processing method for the froth surfaces are special complex grain fully occupied with various sizes of bubbles. A novel visual features extraction and representation method based on the Gabor Wavelet transform is presented in order to classify and identify the flotation production condition in terms of the froth appearance. Firstly, the froth images are filtered by a set of Gabor wavelets of multi-scales and multi-directions; then, the statistical distribution characteristics of the amplitude response representation of the froth image are analyzed by the marginal statistical distribution and joint statistical distribution respectively; at last, the corresponding statistical distribution parameters are extracted and generated as the visual features. A great number of images in various flotation conditions with the production process parameters were collected from a bauxite concentrator. The visual features of the froth images were extracted and the flotation production conditions were classified using the unsupervised fuzzy C-mean clustering analysis. The clustering results demonstrate that this method can extract the distinctive froth visual features and the flotation production conditions can be classified with the visual features effectively.