High-throughput quantification of sensory taste and texture in tomatoes using an industrial NIR sorter: Development of a novel sensory sweet–acidic ratio

Quantifying the sensory attributes of food is traditionally labor-intensive and subjective. This study investigated the feasibility of using an industrial-scale near-infrared (NIR) sorter to non-destructively quantify the taste and texture of tomatoes in high-throughput environments. Two tomato cultivars (‘Round Red’ and ‘High Lycopene’) were evaluated using an NIR sorting system equipped with a silicon sensor (650–950 nm) operating at a speed of six fruits per second. To mitigate the influence of random fruit orientations on the belt conveyors, spectra were acquired from four different positions per fruit. Simultaneously, a rapid sensory evaluation protocol was implemented by five panelists to obtain quantitative scores for five attributes (sweetness, acidic taste, firmness, skin awareness after mastication, and mealiness) using a visual analog scale. Although individual ratings exhibited variability, the average scores provided robust objective variables, leveraging the Law of Large Numbers through high-volume sampling. Calibration models were developed using partial least squares and LASSO regression. After evaluating the models specialized for cultivar and orientation, the global models demonstrated practical estimation performance for grading purposes across most sensory attributes (excluding mealiness) and physicochemical values (Brix and acidity). Furthermore, we proposed a novel index, the “Sensory Sweet–Acidic Ratio” (SSAR), derived from the predicted sensory scores. SSAR exhibited superior stability and correlation compared with traditional sugar–acid ratios, which are often susceptible to seasonal fluctuations. This system achieved a Technology Readiness Level of 8, effectively bridging the gap between laboratory research and industrial application.