Development of an online Nigella sativa inspection system equipped with machine vision technology and artificial neural networks

Aim of study: Nigella sativa L. seeds usually are mixed with impurities, which affect its quality and influences consumer acceptance in both raw seeds and the oil market. In this study, an intelligent system based on the combination of machine vision (MV) and artificial neural networks (ANN) was developed to classify and clean N. sativa seeds and its impurities. Area of study: Iran, Kurdistan province. Material and methods: For accurate detections we developed a robust image processing algorithm including image acquisition, image enhancement, segmentation, and feature extraction steps. Correlation-based Feature Selection method was used to select the superior features. Three methods of linear discriminant analysis, support vector machines, and ANN were used to classify the data. Main results: The statistical indices of sensitivity, specificity, and accuracy for N. sativa in the online phase were 90%, 98.93%, and 97.04%, respectively. The average of these measurements for the impurities class was 95.57%, 96.89%, and 96.58%, respectively. Research highlights: The results demonstrated the feasibility of suggested machine learning and image processing approaches in the real-time cleaning of N. sativa. The image acquisition and processing process, including selection of the best lighting methods to reduce the shadows, noise elimination and segmentation, provided precise results. The final results indicated the effectiveness of proposed machine learning algorithm in feature extraction, feature dimensionality reduction, and classification approaches. This methodology can be recommended for detection, classification and automatic cleaning of other similar seeds.