Fault condition recognition based on multi-scale co-occurrence matrix for copper flotation process

Image processing technology has been successfully applied to fault detection of copper flotation processes, and the key to realize image processing based fault condition recognition is accurately extracting froth image features closely related to key production indices. To extract texture features of froth images in real-time, a multi-scale gray level co-occurrence matrix (M-GLCM) method is proposed. Firstly, the wavelet transform is applied to the froth gray images, and the coefficients of wavelet approximation sub-images in different scales are mapped into gray level of 0~255. Then the spatial gray level co-occurrence matrices are calculated, a set of texture features of froth images are obtained statistically from the matrices. Lastly, these features are adopted for off-line classification and on-line recognition of froth images in different working conditions. As the result shows, froth texture features extracted through the multi-scale gray level co-occurrence matrix have a good stability and separability suitable for mode classification and thus can be well applied for fault condition recognition in copper flotation processes.