Application of handheld near-infrared technology for in-field analysis of non-volatile traits in fresh market tomatoes

This is the authors' abstract. We don't add key points for this paper.

Tomato-breeding companies are continuously developing new cultivars necessitating in-field characterization. Our aim was to develop predictive regression algorithms for rapid characterization (∼10 s) of non-volatile flavor traits using a handheld near-infrared (NIR) device. We collected spectra in reflectance mode on intact tomatoes and transflectance (0.5 mm pathlength) mode for tomato juice. Official methods were employed to determine reference values of quality traits. A large sample size (n > 600) of tomato varieties were analyzed representing diverse variation in parameter concentrations (TA: 0.2 - 1.0 g CA/100ml, citric acid: 0.3 – 1.4%, ascorbic acid: 1.2 - 60.3 mg/100ml, TSS: 2.8 - 9.4 0Brix, glucose: 0.7 % – 2.9 %, and fructose: 0.9% – 3.1%). Partial least square regression (PLSR) was used to generate prediction models. A NIR prototype operating with 3 tungsten lamp showed better performance than its commercial NIR counterpart equipped with 7 tungsten lamps. Models developed using tomato juice showed better performance and underscored their reliability for screening calibration with Rpre ≥ 0.76, low RMSEP, RER ≥ 7.3 and SEP/SEC metric ≤ 1.1. Our findings demonstrate the application of handheld NIR system in transflectance mode for the rapid assessment of multiple quality parameters in fresh tomatoes.