Non-destructive measurement and real-time monitoring of apple hardness during ultrasonic contact drying via portable NIR spectroscopy and machine learning

• Low-cost non-destructive measurement of apple hardness by portable NIRS. Robust and reliable models were developed PLS, MLR ANN. Models based on the ANN had highest prediction performance. Seven featured wavelengths selected to form a compact model. A close-formed equation for online monitoring was developed. Portable near-infrared spectrometer in spectral range 900–1700 nm evaluated first time assess monitor real-time during ultrasonic drying. Calibration using PLS ANN, their performances internal leave-one-out cross-validation an external dataset. Several pre-treatments including standard normal variate (SNV), multiplicative scatter correction (MSC), Savitzky–Golay second derivatives employed examine effects variations prediction. important weighted regression coefficients develop simple model facilitate interpretation circumvent noise. The MLR, with predicted R 2 p 0.91, 0.95, RMSEP 14.78, 14.85, 12.46 N, respectively. results indicate that NIR spectrometers are quite promising