FOOD SCIENCE ›› 2026, Vol. 47 ›› Issue (20): 23-30.doi: 10.7506/spkx1002-6630-20260501-003

• Food Analysis and Detection Based on Spectroscopy and Chemometrics • Previous Articles    

Prediction of Soluble Solids Content and Optimal Harvest Date of Navel Oranges Using Handheld Near-Infrared Spectrometer

HUANG Zhisai, LONG Tao, SUN Xudong, LIU Jiacheng, CHEN Zhaoxing, LÜ Qiang, CHEN Xiaojing, FENG Shaoran, HAN Xi   

  1. 1. School of Mechatronics and Vehicle Engineering, East China Jiaotong University, Nanchang 330013, China; 2. Ganzhou Citrus Research Institute, Ganzhou 341004, China; 3. Citrus Research Institute, Southwest University, Chongqing 400712, China; 4. College of Electrical and Electronic Engineering, Wenzhou University, Wenzhou 325035, China; 5. Beijing Yangguangyishida Technology Co. Ltd., Beijing 100122, China; 6. Beijing Wei Chuang Map Technology Co. Ltd., Beijing 100070, China
  • Published:2026-11-06

Abstract: During the harvest of navel oranges, variations in ambient light in the orchard and inter-instrument differences can lead to a decline in the performance of predictive models based on data from laboratory-calibrated handheld near-infrared spectrometers (NIRS). This study employed a global model and external parameter orthogonalization (EPO) to correct for these interfering factors, thereby enabling the prediction of the optimal harvest time for navel oranges. In this study, 1 352 samples of Gannan navel oranges were collected across two harvest seasons (2023–2024). Spectral data of these samples were acquired and their soluble solids contents (SSC) were measured. The results showed that second-derivative (2D) preprocessing of spectral data in the 700–930 nm range effectively reduced interference from chlorophyll and water. When only inter-instrument differences were present, the global model achieved better cross-instrument prediction performance than the EPO method (R2 = 0.71, root mean square error of prediction = 1.23%). Under the co-existence of interferences from variations in ambient light and inter-instrument differences, incorporating new samples involving both types of differences 收稿日期:2026-05-01 基金项目:江西省国际科技合作项目(20253BDH410018);江西省自然科学基金项目(20253BAC280032;20242BAB25361) 第一作者简介:黄志赛(2001—)(ORCID: 0009-0005-2657-252X),男,硕士研究生,研究方向为近红外光谱。 E-mail: 1263281645@qq.com *通信作者简介:孙旭东(1978—)(ORCID: 0000-0001-6743-1348),男,教授,博士,研究方向为近红外光谱和计算机 视觉果业应用。E-mail: sunxudong\_18@163.com into the global model at 15% of the calibration set yielded the best corrective performance with R2 of 0.59 and 0.53 for indoor and orchard testing, respectively. Using the calibrated model to predict the SSC of eight batches of oranges harvested in 2024, we found that the proportion of samples with SSC ≥ 12% increased significantly in oranges harvested on November 26th, with the highest predicted value exceeding 95%, indicating that November 26th is the optimal harvest date for Gannan navel oranges. In summary, the global model effectively overcomes spectral interference caused by inter-instrument and cross-ambient light conditions, providing reliable methodological support for rapidly predicting the SSC and determining the optimal harvest date of navel oranges in orchard environments.

Key words: navel orange; quality; optimal harvest date; near-infrared spectroscopy; hand-held

CLC Number: