1. No category

A Novel Approach of Co­Occurrence Matrix for Gangue Separation from Iron Debi Prasad Tripathy Professor, Department of Mining Engineering, National Institute of Technology, Rourkela, India +91­661­2462608 [email protected] K. Guru Raghavendra Reddy Research Scholar, Department of Mining Engineering, National Institute of Technology, Rourkela, India +91­7853834547 [email protected] ABSTRACT Ore sorting is a useful tool to remove gangue material from the ore and increase the quality of the ore. The vast developments in the area of artificial intelligence allows fast processing of full color digital images for the preferred investigations. In this paper, a novel approach to categorize the ores of iron feed has been proposed based on analyzing color­texture features using two different approaches, based on extensions of the co­occurrence matrix method. In the first method, co­occurrence matrices were computed both between and within the color bands and the second method used joint color­texture features. The dimensions of the image features were reduced by applying Euclidean Distance ranking. A multi­layer perceptron neural network model was used as a mapping function to classify the material. 1.
INTRODUCTION Automatic sensor­based ore sorting (Salter and Wyatt, 1991; De Jong et al., 2004) is a major breakthrough in minerals technology and upfront beneficiation resulting in substantial reduction in downstream costs, improvement of ore quality and exploitation of low grade ore reserves. Digital image processing techniques in recent times have been applied in the mineral industry in different mining operations such as online ore monitoring, particle size estimation, ore sorting and classification. Image processing is a potential approach to develop an online system, which stimulates human eyes to detect gangue mineral particles in the ore based on the visual textural properties of ores. This study provides an innovative approach from discrimination of iron ore from associated gangue mineral using image analysis based on color and texture features. 2.
METHODOLOGY The main steps involved in the proposed method are: image collection, image pre­processing, image segmentation, feature selection and image recognition. These steps are described in the following sub­sections (see Figure 1) (Wang and Liang, 2011). Figure 1.​
​
Image processing based ore classification method 3.
EXPERIMENT AND RESULTS Samples were collected from TRB Mines, Odisha in India. The images of the samples collected were taken in the laboratory set up with illumination system. The size of the imported jpeg images was 2MB. The images acquired were 4000×3000 pixels in size. Images of different types of samples that were collected (See Figure 2). Figure 2.​
​
Images of laterite, hematite, limonite and shale samples The segmented images were then processed for identifying and labelling the individual rock present in the segmented parts using regional labelling algorithm. The network was trained using 70% of the data and 30 % for validation and testing. The results indicated 86.7% accuracy. 4.
CONCLUSION A vision­based ore sorting model based on analysis of color­texture features is presented. Initially the color features were computed both between and within the color bands of RGB HSV and YCbCr color spaces. The study results indicated the effectiveness of MLP neural network was used for identifying gangue minerals. This study results indicated the effectiveness of the above vision­based gangue separation was 86.7% accurate. 5.
REFERENCES Al­Batah, M.S., Isa, N.A.M., Zamli, K.Z., Sani, Z.M. and Azizli, K.A. 2009. A novel aggregate classification technique using moment invariants and cascaded multilayered perceptron network. ​
International Journal of Mineral Processing​
. 92, 1, 92­102. Arvis, V., Debain, C., Berducat, M. and Benassi, A. 2011. Generalization of the cooccurrence matrix for colour images: application to colour texture classification. ​
Image Analysis & Stereology​
. 23, 1, 63­72. Chatterjee, S. 2013. Vision­based rock­type classification of limestone using multi­class support vector machine. ​
Applied Intelligence​
. 39, 1, 14­27. Chatterjee, S., Bandopadhyay, S. and Machuca, D. 2010. Ore Grade Prediction Using a Genetic Algorithm and Clustering Based Ensemble Neural Network Model. ​
Mathematical Geosciences​
. 42, 3, 309­326. Chatterjee, S., Bhattacherjee, A., Samanta, B. and Pal, SK. 2010. Image­based quality monitoring system of limestone ore grades. ​
Computers in Industry​
. 61, 5, 391–408. Conners, RW. and Harlow, CA. 1980. A theoretical comparison of texture algorithms. ​
IEEE Transactions on Pattern Analysis and Machine Intelligence​
. 2, 3, 204­22. Das, S. and Choudhury, M.R. 2014. Rock type classification by image analysis using the quaternion colour extraction model and support vector machine classifier. Journal of Oil, Gas and Coal Engineering​
. 1, 1, 002­009. De Jong, T.P., van Houwelingen, J.A. and Kuilman, W. 2004. Automatic sorting and control in solid fuel processing: opportunities in European perspective. Geologica belgica​
. 7, 3, 325­333. Drimbarean, A. and Whelan, P.F. 2001. Experiments in colour texture analysis. Pattern recognition letters​
. 22, 10, 1161­1167. Gao, K., Du, C., Wang, H. and Zhang, S. 2013. An Efficient of Coal and Gangue Recognition Algorithm. ​
International Journal of Signal Processing, ​
Image Processing & Pattern Recognition​
. 6, 4,345­354. Gonzalez, R.C., Woods, R.E. and Eddins, S. L. 2004. ​
Digital image processing using MATLAB​
. Pearson Education India. Khorram, F., Memarian, H., Tokhmechi, B. and Soltanian­zadeh, H. 2011. Limestone chemical components estimation using image processing and pattern recognition techniques. ​
Journal of Mining & Environment​
. 2, 2, 49­58. Li, W., Wang, Y., Fu, B. and Lin, Y. 2010. Coal and Coal Gangue Separation Based on Computer Vision. ​
Fifth International Conference on Frontier of Computer Science and Technology.​
467­ 472. Zhang, Z., Yang, J., Ding, L. and Zhao, Y. 2012.Estimation of coal particle size distribution by image segmentation. ​
International Journal of Mining Science and Technology​
. 22, 5, 739­744.