Shima Farahbakhsh - M. Sc Student in Shahid Bahonar University of Kerman, Iran

Nowadays, most industrial practices have replaced human resources with mechanical or robot workers. Like industry, the importance of applying automated systems is also strongly felt in agriculture practices. This study proposes two-phase model for identifying ripe walnuts from unripe ones; during the first phase of this model, walnuts and non-walnuts are separated using the proposed model which is based on local binary pattern (LBP) approach. This phase uses Center symmetric Local Binary Pattern (CS-LBP) in HSV color space to extract features and uses Support Vector Machine (SVM) for classification and training. To compare results, in addition to the Support Vector Machine, 3-level multilayer perceptron (MLP) neural network was also used for classification and training. Then, during the second phase of the proposed method set of basic principles were extracted with regard to experts’ opinions and, after that, walnuts were first classified into two groups of ripe and unripe using Support Vector Machine and again they were classified into three groups from unripe to ripe by means of the proposed method based on Fuzzy system, decision tree and the extracted Fuzzy principles; in the end, methods were evaluated according to their accuracy and recognition abilities. Results show that the average accuracy of Support Vector Machine method after implementations is equal to 84.28 and walnuts can only be classified into two groups of unripe and ripe However, the proposed method includes three groups of unripe half-ripe and ripe and its average accuracy after application is 88.33.

Fuzzy logic, image processing, image texture, ripe walnut, recognition;