Separation of walnut shell from kernels and kernel classification based on color using invariable moments, artificial networks and the method of discriminant analysis

Introducing a method for separating the shell from kernel by considering the high price difference between kernels is very important from standpoint of color (light to amber). Utilization of the developed method can reduce the requirements of labor force and subsequently will be economically useful. In this purpose، a research was done on Persian walnut (Juglans regia L.). In this study، machine vision technology accompanying with artificial networks were used to study fundamental principals which can be applied for automating the process of separation shell from kernels and kernel classification. Mentioned method includes two phases: separation shell from kernel and classification of kernel. Five invariant moments were applied for separating shell from kernel. After calculating moments an artificial network was used to separate two groups. The results showed that the correct separation rate in the first phase is about 98. 6%. In the second phase، to categories walnut kernels، four categories were considered (extra light، light، light amber and amber). By comparing two color spaces HSI and RGB with applying two principal components R and G (RGB) and using discriminant analysis، the process accuracy was obtained about 98. 15%. The explained method can be a proper way for classification of walnut kernels from light to amber color.