Based On Color Co-occurrence Matrix Texture Feature Extraction And Application

Froth flotation is the most widely used method to extract certain minerals from ore. During the flotation process, foam layer can be a direct response to the quality indicators. At present, the flotation operations in the concentrator are usually adjusted by experienced workers through observing froth surface condition. However, these types of flotation process are very difficult to control optimally owing to the limitations of time, space and subjectivity. Therefore, in order to achieve optimal control of mineral flotation process based on digital image processing, the method to extract the texture features of froth images is studied and used to guide the flotation industrial process operation primarily in this paper.Based on the analyses of the mechanism of flotation froth forming, the relationship between the froth surface visual feature and flotation conditions is studied firstly in this paper. Considering the limitations of gray level co-occurrence matrix(GLCM) applied to the color bubble images, a new texture extraction algorithm based on color co-occurrence matrix(CCM) is proposed by combining the characteristics of froth image. First, the color space for froth image is converted and each color components is quantified. Then, the color co-occurrence matrix is calculated for the froth image, and the feature statistics is extracted from the normalized co-occurrence matrix. Finally, according to the froth texture condition represented by the obtained feature statistics, the new froth texture complexity parameter is designed. By analyzing the relativity between the extracted froth texture complexity and the grade of mineral, the effectiveness of the proposed method is proved and optimum texture complexity range is given.In view of the lack of accuracy for the existing flotation condition classification, more accurate classification for froth condition is obtained by using foam texture features extracted from CCM. Bubble characteristic parameters are analyzed by Principle Component Analysis method firstly, and then are clustered by using Fuzzy C-Means method. By studying clustering evaluation criteria, the experimental comparison and analysis vividly demonstrates that the flotation conditions can be well recognized and achieve effective clustering through characteristic parameters extracted by CCM, which provides important guidance for the flotation production operation.