Detecting Hard-to-Cook Chickpeas with Hyperspectral Imaging

Saha, D., Senthilkumar, T., Singh, C. B., Pauls, P., & Manickavasagan, A. (2023). Rapid and non-destructive detection of hard to cook chickpeas using NIR hyperspectral imaging and machine learning. Food and Bioproducts Processing, 141, 91–106. https://doi.org/10.1016/j.fbp.2023.07.006

This study explores the use of Near-Infrared Hyperspectral Imaging (NIR-HSI) for the rapid and non-destructive classification of chickpeas as either Hard to Cook (HTC) or Easy to Cook (ETC). Researchers created two types of HTC chickpeas—through suboptimal storage and chemical treatment—using eight chickpea varieties. Over 864 seeds were analyzed using hyperspectral imaging within the 900–2500 nm spectral range.

To correlate image data with cooking behavior, the cooking times were measured using an automated Mattson cooker. Advanced mathematical models, including Partial Least Squares Discriminant Analysis (PLSDA), Support Vector Classifier (SVC), and a Convolutional Neural Network with Attention (CNN-ATT), were used to classify chickpeas. Both SVC and CNN-ATT models achieved 100% classification accuracy.