Limestone chemical components estimation using image processing and pattern recognition techniques

In this study, an ore grade estimation model was developed based on image processing and pattern recognition techniques. The study was performed at a limestone mine in central part of Iran. The samples were randomly collected from different parts of the mine and crushed down (from 10 cm to 2.58 cm). The images of the samples were taken in an appropriate environment and processed. A total of 76 features were extracted from the identified rock samples in all images. Neural network was used as an intelligent tool for ore grade estimation. First, six principal components derived from principal component analysis were used as input of neural network and four grade attributes of limestone (CaCO3, Al2O3, Fe2O3 and MgCO3) were used as the output. The root of mean squared error between the observed values and the model estimated values for the test data set were 0.38, 0.84, 0.15 and 0.03; the R2 values were 0.78, 0.76, 0.76 and 0.81 for the mentioned chemical compositions respectively. The value of R2 indicates the correlation between the actual and estimated data. It can therefore be inferred that the model could successfully estimate the percentage of chemical compositions of the samples collected from the same mine.