A classification system for beans using computer vision system and artificial neural networks

A computer vision system (CVS) was developed for the quality inspection of beans, based on size and color quantification of samples. The system consisted of a hardware and a software. The hardware was developed to capture a standard image from the samples. The software was coded in Matlab for segmentation, morphological operation and color quantification of the samples. For practical application of the software, a user-friendly interface was designed using Matlab graphical user interface (GUI). Length and width of the samples were determined using this system. Then the results of the system were compared to the measurements obtained by a caliper. High correlations (r = 0.984 and 0.971 for length and width, respectively) were obtained between the results of the system and the caliper measurements. Moment analysis was performed to identify the beans based on their intensity distribution. Average, variance, skewness and kurtosis values were determined for each channel of RGB color format. Artificial neural networks (ANN) were used for color quantification of the samples. Samples were classified by human inspectors into five classes and twelve moment values of the 69 samples with their classes were used in the training stage of ANN. Testing of the ANN was performed with other 371 samples. The automated system was able to correctly classify 99.3% of white beans, 93.3% of yellow–green damaged beans, 69.1% of black damaged beans, 74.5% of low damaged beans and 93.8% of highly damaged beans. The overall correct classification rate obtained was 90.6%.