Classification of seed grains on the basis of their surface pattern

Abstract

Classification of valuable seeds and identification of foreign materials are very important in optimization of cleaning process and plant development. Digital image processing was applied to measure statistical parameters of surface patterns, such as mean, entropy, angular second momentum and contrast. In addition, polar quality points of histograms of intensity differences were also computed. A distance function of dimensionless quantity was introduced and applied in classification. Distances were calculated from average value and standard deviation of parameters. Advantages of this method are the easy calculation and assignment of probabilities to distance values. Not only the grey level differences were collected, the method was extended to colour signals (red, green and blue) as well. Analysis of colour information improved effectiveness of classification.