This study aims to develop rapid and non-invasive methods based on near-infrared hyperspectral imaging and chemometrics for quantitative prediction of chemical compositions of pea-derived products. Hyperspectral imaging was used to acquire images from pea processing streams, namely pea flour, pea protein concentrate, and pea protein isolate. The PLS algorithm was used to develop quantitative prediction models based on the relationship between the hyperspectral image data and the chemical compositions of the pea products, including moisture, protein, ash, insoluble fiber, and total starch. Prediction results in terms of coefficient of determination (R2) and root mean square errors in the prediction (RMSEP) datasets show accurate results for moisture (R2=0.844, RMSEP = 0.407%), protein (R2=0.99, RMSEP = 2.074%), ash (R2=0.778, RMSEP = 0.474%), and total starch (R2=0.991, RMSEP = 2.316%) contents. Low prediction accuracy was obtained for insoluble fiber (R2=0.597, RMSEP 2.474%) content. The accurate prediction achieved by hyperspectral imaging highlights its suitability for high throughput multi-parameter assessment of pea-derived products, which is particularly important given their increasing market demand.
Read full abstract