Grain size measurement using distribution classification

The concept of grain size distribution is important in many industrial processes. in this study, rolled chrome concentrate was investigated to find the average grain size. Distribution classification was used with covariance, gray level difference, LBP, Laws and color features. As a comparison cooccurrence features were used. Classification was performed by a distribution classifier using leave-1-out and holdout methods. Test material was sieved into 15 fractions, from 37 mu m to 500 mu m, and also mixtures of three adjacent fractions were formed. The best results were obtained with gray level difference features that could classify 15 fractions without error. Finally grain mixtures of three adjacent fractions were analyzed using the holdout method. The results show that the combination of gray level difference features and distribution classifier can distinguish test material quite well.