Prediction of soluble-solid content in citrus fruit using visible–near infrared hyperspectral imaging based on machine learning and effective-wavelength selection algorithm

Authors: KIM, MON-JEE - Kangwon National University; YU, WOO-YOUNG - Kangwon National University; SONG, DOO-JIN - Kangwon National University; CHUN, SEUNG-WOO - Kangwon National University; Kim, Moon; LEE, AHYEONG - Korean Rural Development Administration; KIM, GIYOUNG - Korean Rural Development Administration; MO, CHANGYEUN - Kangwon National University

Submitted to: Sensors
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/20/2024
Publication Date: 2/26/2024
Citation: Kim, M., Yu, W., Song, D., Chun, S., Kim, M.S., Lee, A., Kim, G., Mo, C. 2024. Prediction of soluble-solid content in citrus fruit using visible–near infrared hyperspectral imaging based on machine learning and effective-wavelength selection algorithm. Sensors. 24, 1512. https://doi.org/10.3390/s24051512.
DOI: https://doi.org/10.3390/s24051512

Interpretive Summary: The soluble solids content (SSC) of a fruit is an important quality measure highly correlated to ripeness and sweetness. Conventional methods of measuring SSC are sample-destructive and time-consuming, and therefore unsuitable for assessing high volumes of fruit to be sold intact. Hyperspectral imaging techniques can often provide a non-destructive means to evaluate attributes beyond a fruit surface. This study used hyperspectral imaging methods in the visible/near-infrared (VNIR) region to develop SSC prediction models for citrus based on partial least-squares regression (PLSR). Hyperspectral VNIR imaging of 324 intact fruits of one citrus cultivar was conducted to acquire calyx-end and blossom-end images. Immediately after each fruit was imaged, its pulp was juiced for SSC measurement with a digital refractometer to acquire Brix values for use in model development, in conjunction with a variety of spectral image preprocessing techniques. A PLSR model incorporating competitive adaptive reweighted sampling in the spectral preprocessing showed promising results for nondestructive imaging-based prediction of SSC, and will be used for further studies incorporating a greater variety of citrus cultivars and developing an online system for SSC evaluation of the fruits for use in post-harvest agricultural management and processing for fruit producers or distributors.

Technical Abstract: Citrus fruits are sorted based on external qualities, such as size, weight, and color, and internal qualities, such as soluble-solid content (SSC), acidity, and firmness. Visible and near-infrared (VNIR) hyperspectral-imaging techniques serve as rapid and nondestructive techniques for de-termining the internal quality of fruits. This study evaluates the applicability of the VNIR hy-perspectral-imaging technique for predicting the SSC in citrus fruits. A VNIR hyperspec-tral-imaging system with a wavelength range of 400–1000 nm and 100 W light source is used to acquire hyperspectral images from citrus fruits in two orientations (i.e., stem and calyx ends). The SSC-prediction model is developed using partial least-squares regression (PLSR). Spectrum pre-processing, effective-wavelength selection through competitive adaptive reweighted sampling (CARS), and outlier detection are used to improve model performance. The performance of each model is evaluated using the coefficient of determination (R2) and root mean square error (RMSE). The SSC-prediction CARS-PLSR model with outliers removed exhibited R2 and RMSE values of approximatively 0.75 and 0.56 °Brix, respectively. The results of this study are expected to be useful in similar fields such as agricultural and food post-harvest management as well as in the development of an online system for determining the SSC of citrus fruits.