Identification and classification of persian Cumin (Bunium persicum Boiss) landraces using image processing in combination with artificial neural networks

Persian cumin (Bunium persicum) is one of the medicinal plants with high economic and expert value which its identification and classification is very important. Manual or visual inspection and classification of Persian cumin is very time-consuming and associated with errors. Therefore machine vision technology as a novel and non-destructive method can be a very good technique for identification and classification of the product. The aim of this study was to identify and to classify Persian cumin landraces based on color and texture features using image processing and artificial neural network. In this research, seven samples of Persian cumin landraces were collected from natural habitats of Kerman province and after image acquisition of samples, 36 color features and 108 textural features were extracted from the images. Identification of landraces was carried out using back propagation ANNs. Based on the results of the study, the mean classification accuracy using a one layer ANN, for the color, texture and color-texture features was equal to 93.55%, 93.50% and 96.40%, respectively. Also, the minimum value of mean square error for color, texture and color-texture features were obtained 0.172, 0.182 and 0.148, respectively.