Firmness measurement of kiwifruit using a self-designed device based on acoustic vibration technology

Firmness is an important quality indicator of kiwifruit and is useful for determining optimal time for marketing and optimizing storage management. In this study, a self-designed system based on acoustic vibration technology was used to realize real-time detection of kiwifruit firmness. To ensure stability of measurement, impact force of the excitation device in the system was calibrated to 12.03 +/- 0.71 N. The acoustic vibration response signals of kiwifruit were converted from time domain to frequency domain and 10 statistical features were extracted. Most of the features had good correlations with reference firmness, which proved feasibility of firmness prediction using signals collected in the self-designed system. Subsequently, PLS regression models for predicting firmness were established based on frequency-domain spectra. To further improve accuracy of the model, CARS algorithm was used to select effective frequencies that were highly correlated with kiwifruit firmness. According to the results, prediction accuracy of the CARS-PLS model in external cross-validation sets for flesh firmness was the best (R-cv(2) = 0.96, RMSECV = 0.27, and RPDcv = 5.21), followed by stiffness (R-cv(2) = 0.95, RMSECV = 0.43, and RPDcv = 5.00), and prediction accuracy of the model for skin firmness was the worst (R-cv(2) = 0.93, RMSECV = 0.81, and RPDcv = 4.01). Overall, acoustic vibration signals obtained by the self-designed device in a nondestructive way can well characterize the firmness of kiwifruit. The proposed method in this study can achieve high-precision prediction of all three kiwifruit firmness indices and meet requirements of online real-time detection.