Prediction of protein content in paddy rice (Oryza sativa L.) combining near-infrared spectroscopy and deep-learning algorithm

• DOI: 10.3389/fpls.2024.1398762
• Publication Date: 2024-07-31T05:14:13Z
• Authors: Ha-Eun Yang; Nam-Wook Kim; Hong-Gu Lee; Min-Jee Kim; Wan-Gyu Sang; Changju Yang; Changyeun Mo
• Keywords: Rice protein; Brown rice; Husk

Rice is a staple crop in Asia, with more than 400 million tons consumed annually worldwide. The protein content of rice major determinant its unique structural, physical, and nutritional properties. Chemical analysis, traditional method for measuring rice’s content, demands considerable manpower, time, costs, including preprocessing such as removing the husk. Therefore, technology needed to rapidly nondestructively measure paddy during harvest storage stages. In this study, nondestructive technique predicting husks (paddy rice) was developed using near-infrared spectroscopy deep learning techniques. prediction model based on partial least square regression, support vector neural network (DNN) were spectrum range 950 2200 nm. 1800 spectra 1200 from brown obtained, these used development performance evaluation model. Various spectral techniques applied. DNN showed best results among three types models. optimal first-order derivative accuracy coefficient determination prediction, R p 2 = 0.972 root mean squared error RMSEP 0.048%. applied 0.987 0.033%. These demonstrate commercial feasibility non-destructive both husked seeds rice.