Sulfur Flotation Froth Image Dynamic Texture Modeling Applying On Recognition Of Conditions

In the flotation of Sulfur, the flotation performance is reflected by the surface visual features. The boundary among Sulfur flotation froth is obvious, but the amount of foam with mineral unstable, the size of the bubble and its change is random. The normal surface characteristics of foam is difficult to reflect the flotation condition accurately, so texture become a key feature of characterization of the surface state of a sulfur froth. Currently operating in the production of sulfur flotation mainly rely on manual adjustment of the variable observed foam texture features, but due to the limitations of human observation, it is difficult to ensure flotation in the best condition. Therefore, the use of digital image processing technology is reasonable for the analysis of sulfur flotation froth image characteristics, sulfur flotation froth image texture feature extraction methods, the selection of foam texture quantitative guidance flotation of great significance and engineering application value.Since sulfur concentrate is zinc concentrate slurry after milling and slurry formed by acid leach residue after leaching the flotation circuit, the resulting pulp flotation separation. In actual production, sulfur flotation conditions change quickly, often within a variety of abnormal conditions in a short period, affecting the flotation efficiency seriously. Therefore, a single bubble image features can not accurately reflect the flotation condition is good or bad. For the Characteristics of sulfur froth, we first study the various types of conditions of the sulfur flotation, and the corresponding bubble image, flotation froth image is divided into six categories in accordance with the conditions, as the standard of the subsequent classification and recognition. Using gray level co-occurrence matrix and gray-gradient co-occurrence matrix to extract static bubble image texture features and the static texture features for classification and identification.For single-frame froth image features can not reflect the flotation conditions accurately, the proposed use of the ARMA (autoregressive moving average) dynamic texture model to describe the correlation between the bubble image. Training model parameters A, C, Q, R by Sample learning, and then synthesize sulfur flotation froth image dynamic texture and the synthesis results to judge the good and bad effects of parameter learning of model parameters.Finally, the sample static texture features as the input feature vector of the support vector machines for classification and identification of sulfur froth. The modified model parameters A, C and Martin distance for classification and identification of dynamic texture image sequence, This will not only make up a single frame image features the shortcomings but can seize the dynamic characteristics of the image sequence, and then comparing Dynamic texture recognition and static texture recognition.