Development and evaluation of chickpea classification system based on visible image processing technology and artificial neural network

The ability to recognize color, texture, and shape in image processing technology has led to the development of machine vision systems in various fields of agriculture, conversion industries, and industry. The existence of impurities and seeds with an unsuitable appearance in chickpeas and the direct effect of the appearance quality of the product on its marketability, shows the need for grading this product. The aim of the present study was to distinguish impurities and chickpeas with an inappropriate appearance from chickpeas with a suitable appearance by developing a machine vision system. Totally 270 images including 54 samples of chickpeas with suitable appearance and 36 samples of each type of chickpeas with inappropriate appearance (wrinkled, green, brown, and split) and foreign materials (stone and stem) were prepared. After preparing the sample images, the pre-processing and feature extraction steps were performed automatically and different color, texture and shape properties were extracted by developing and using an image processing algorithm. An algorithm was developed to select efficient features from the extracted features. Efficient features were classified by the artificial neural network method with total accuracy of 91.9%. The detection accuracy for desirable, wrinkled, cotyledon, immature, and brown chickpea and stem and stone impurities was equal to 98.1, 83.3, 100.0, 91.7, 97.2, 77.8, and 97.2% respectively. Using the developed system, the chickpea product can be graded with high accuracy and low cost, so that after separating the impurities, the desirable and undesirable chickpeas can be separated and sent to the market for different uses.