Quality Assessment of Grain Samples using Color Image Analysis

Mirolyub Mladenov, Stanislav Penchev, Martin Dejanov, and Metin Mustafa


grain quality assessment, feature extraction, classification, computer vision


Grain quality is assessed on the basis of different characteristics like appearance, shape, color, infections, presence of impurities etc. Most of these characteristics are assessed by an expert using visual estimation. In this paper, an approach for objective estimation of some basic grain quality characteristics based on color image analysis of the investigated objects is presented. Methods and tools for feature extraction and for object description, as well as for object classification into preliminary defined groups are proposed. Three classification procedures based on radial basis elements are presented. The possibility of using them to solve different classification problems is analysed and its training and validation accuracy is assessed. The results from the validation procedure give opportunity to choose appropriate data model, classifier and some of classifier parameters for specific classification task. Results from the classification of maize grain samples, which include objects from 9 classes with different color characteristics and 4 classes with different shape are presented. The final classification of the objects is performed in 3 normative classes by fusing the data from color characteristics and shape classification.

Important Links:

Go Back