Estimation of copper titration or calibration from flotation floor using machine image analysis and vision

By observing froth surface, operators can usually determine the metallurgical parameters (grade and recovery) of flotation process but it is associated with many problems such as inability to continuously monitor, significant differences with operation results, and various observations by different persons. In this study, the physical and structure properties of flotation froth images have been investigated to determine the metallurgical parameters of copper oxide ore. Pre-processing and processing of images obtained from the flotation froth was studied using artificial neural networks in MATLAB program. The estimated grade was compared with the actual grade to check the accuracy of the system. Studies show that when the three color characteristics (red, green, and blue colors) of images are used to determine the grade of froth, these three characteristics alone are not sufficient to estimate the grade, and the amount of error rate is 21.7%. When factors of the three color characteristics, and their standard deviation was studied for estimating the grade, the error rate reached 8.7%. Finally, the simultaneous studies of 11 characteristics including color channels, their standard deviation and haralic characteristics (entropy, contrast, energy, correlation, and gradient-density) showed that a very good similarity between the actual grade and their predicted grade. The calculated error rate greatly reduced to 2.3%. The method of work was including of the picture taking of froth flotation, pre-processing and processing of images, characteristics extraction, system training, testing and validation, and finally the results showed that the amount of froth grade in the online form.