He Hong-Ju, Wang Yuling, Wang Yangyang, Liu Hongjie, Zhang Mian, Ou Xingqi
School of Food Science, Henan Institute of Science and Technology, Xinxiang 453003, China.
School of Chemistry, Chemical Engineering and Biotechnology, Nanyang Technological University, Singapore 637459, Singapore.
Food Chem X. 2023 Mar 8;18:100631. doi: 10.1016/j.fochx.2023.100631. eCollection 2023 Jun 30.
This study aimed to achieve the rapid evaluation of moisture, ash and protein of sweet potato simultaneously by near-infrared (NIR) hyperspectral imaging (900-1700 nm). Hyperspectral images of 300 samples for each parameter were acquired and the spectra within images were extracted, averaged and preprocessed to relate to the three measured parameters, using partial least squares (PLS) algorithm, respectively, resulting in good performances. Nine, eleven and eleven informative wavelengths were selected to accelerate the prediction of the three parameters, generating a correlation coefficient of prediction ( ) of 0.984, 0.905, 0.935 and root mean square error of prediction (RMSE) of 0.907%, 0.138%, 0.0941% for moisture, ash and protein, respectively. By transferring the best optimized PLS models to generate color chemical maps, the distributions and variations of the three parameters were visualized. NIR hyperspectral imaging is promising and can be applied to simultaneously evaluate multiple quality parameters of sweet potato.
本研究旨在通过近红外（NIR）高光谱成像（900 - 1700纳米）实现对甘薯水分、灰分和蛋白质的快速同时评估。采集了每个参数300个样本的高光谱图像，并提取、平均和预处理图像中的光谱，分别使用偏最小二乘法（PLS）算法将其与三个测量参数相关联，结果表现良好。选择了9个、11个和11个信息波长以加速对这三个参数的预测，水分、灰分和蛋白质的预测相关系数（ ）分别为0.984、0.905、0.935，预测均方根误差（RMSE）分别为0.907%、0.138%、0.0941%。通过转移最佳优化的PLS模型生成彩色化学图，可视化了这三个参数的分布和变化。近红外高光谱成像具有前景，可用于同时评估甘薯的多个品质参数。
