Postharvest kiwifruit continues to ripen for a period until it reaches the optimal “eating ripe” stage. Without damaging the fruit, it is very difficult to identify the ripeness of postharvest kiwifruit by conventional means. In this study, an electronic nose (E-nose) with 10 metal oxide semiconductor (MOS) gas sensors was used to predict the ripeness of postharvest kiwifruit. Three different feature extraction methods (the max/min values, the difference values and the 70th s values) were employed to discriminate kiwifruit at different ripening times by linear discriminant analysis (LDA), and results showed that the 70th s values method had the best performance in discriminating kiwifruit at different ripening stages, obtaining a 100% original accuracy rate and a 99.4% cross-validation accuracy rate. Partial least squares regression (PLSR), support vector machine (SVM) and random forest (RF) were employed to build prediction models for overall ripeness, soluble solids content (SSC) and firmness. The regression results showed that the RF algorithm had the best performance in predicting the ripeness indexes of postharvest kiwifruit compared with PLSR and SVM, which illustrated that the E-nose data had high correlations with overall ripeness (training: R² = 0.9928; testing: R² = 0.9928), SSC (training: R² = 0.9749; testing: R² = 0.9143) and firmness (training: R² = 0.9814; testing: R² = 0.9290). This study demonstrated that E-nose could be a comprehensive approach to predict the ripeness of postharvest kiwifruit through aroma volatiles.

Inadequate drying and overdrying of grains due to the absence of timely and reliable information on moisture content (MC) aggravates grain postharvest loss. In this study, we investigate the real-time acquisition of MC using microwave free-space transmission measurements. A multifrequency swept signal is used to acquire the attenuation and phase-shift spectrum (2.00–10.00 GHz) of the paddy samples (9.553%–29.633% w.b.) for six thicknesses (1–6 cm). Based on the four rules that are followed in the phase-shift measurement, a phase-shift correction algorithm without any restrictions on the sample thickness is proposed to solve the phase-shift ambiguity that occurs when the sample thickness exceeds the wavelength. To choose the most effective frequencies, 17 candidate frequency subsets are generated by the recursive feature elimination algorithm, and the optimal frequency set, containing eight individual frequencies, is selected by using the Friedman and Nemenyi post hoc tests. The sample thickness and microwave characteristics are used as the input variables to establish the prediction models to achieve a thickness-independent measurement of the MC, and the support vector machine model yielded the best performance ( $R^2$ = 0.992, RMSE = 0.555%, and MAE = 0.398%). The results of this study should encourage future research on the real-time acquisition of reliable MC information in food processing and agriculture-related industries.

In this study, an on-line microwave moisture sensing system (OM2S2) based on multi-frequency swept technique was developed to monitor the moisture content (MC) of corn in the fresh to dry state (MC ranged from 10.89% to 63.64%) in real time. Attenuation and phase-shift data were collected under a frequency swept signal containing 801 frequencies from 2.00 to 10.00 GHz with a 10 MHz interval. For removing the inefficient frequencies, the optimized frequencies were selected by a two-stage frequency selection framework: (1) 17 frequency subsets were generated by random forest-recursive feature elimination algorithm, and then (2) the optimal frequency set (including eight individual frequencies) was determined by voting strategies according to the results of 10-fold cross-validation. The attenuation and phase-shift data corresponding to the optimal frequency set was utilized as the input variables of six machine learning algorithms for MC prediction. A deep neural network (R2=0.997, RMSE=1.087, MAE=0.868) performed best according to Friedman test and Nemenyi post-hoc test and deployed to OM2S2. These results showed that OM2S2 was capable of measuring the MC of corn from fresh to dry state in real time, and it also exhibited potential usage for on-line determination of high MC in food processing and agriculture-related industries.