| Ripeness is an important factor that contribute to the quality of fruits during their storage and shelf life and is one of the main factors that affect conusmers' purchase. It has important significance for accurate detection of fruit ripeness for postharvest storage, transportation, sales and consumption etc.. Nowadays, a variety of nondestructive detection technologies have provided strong technical support for the rapid, non-destructive and accurate assessment of watermelon ripeness. Among these nondestructive detection technologies, detection of the ripeness of watermelon based on the acoustic vibration detection technology has been widely concerned by researchers at home and abroad. Laser Doppler vibrometry (LDV) is a kind of non-contact acoustic vibration detection technology, which has many advantages, mainly including high sensitivity, fast dynamic response, insensitive to the transverse vibration interference. LDV can measure the vibration of fruits accurately, so that this technology could identify the ripeness of fruits. In this research, nondestructive detection of the ripeness of watermelon based on laser Doppler vibrometer technology, exploring the feasibility of applying the LDV technology to the detection of watermelon internal hollow heart. Main research contents, results and conclusions are summarized as follows:(1) The repeatability of the laser Doppler vibrometer system was analyzed and the effects of detection parameters on vibration signals of watermelon were investigated. The the statistical results of difference coefficient (D) show that the D values of samples are less than 50, the laser Doppler vibrometer system has satisfied repeatability. The results of single factor experiment show that the influence of the amplitude of acceleration and the frequency sweep rate are significant on the vibration response signals at the a=0.01, and the effect of the testing point on the vibration response signals is not significant. The results of the interaction of multi-factor orthogonal experiment show that the optimal combination of the amplitude of acceleration, the frequency sweep rate and for watermelon vibration measurement is 2.5g,1000 Hz/min and the sunny side of the equator, respectively.(2) The changes of ripeness indicators that includes sugar content, effective acidity(pH), color value (Lab) and modulus of elasticity were analyzed with different ripeness of watermelon. The results showe that with the maturing of watermelon, the sugar content increases gradually, the pH value and elastic modulus increase first and then decrease, and the color value tends to decrease and then increases.(3) The characteristic parameters of the vibration response signals of watermelon were extracted and the law of the variation of the characteristic parameters were analyzed with the ripeness of watermelon increase.12 characteristic parameters were extracted from the amplitude spectrum of the watermelon vibration response signal:the first, second and triple resonance frequency (f1, f2, f3), the amplitudes of f1, f2, f3 (A1, A2, A3), the first, second and triple stiffness coefficient (S1= f12m, S2=f22m, S3= f32m), and 3 ratios of resonance frequencies (f21= f2/f1, f32=f3/f2, f31=f3/f1). With the ripeness of watermelon increase, f1, f2, f3 decrease gradually, A1, A2, A3 of unripe and ripe watermelons are higher than that of over-ripe watermelons. The f21, f32, f31 gradually increase. S1 tends to increase and then has a decrease tendacy, while S2 and S3 increase gradually.(3) Using multiple linear regression (MLR) and partial least squares regression (PLSR) method to establish the quantitative regression models of ripeness indicators of watermelon based on the factors of vibration characteristic parametersand sample mass (m). The results show that:1) The results of MLR model of sugar content are betetr, rc is 0.77, RMSEC is 1.77°Brix, rp is 0.67 and RMSEP is 1.52°Brix.2) The results of PLSR mogel of pH are better, rc is 0.59, RMSEC is 0.33, rp is 0.54, and RMSEP is 0.34.3) For color values a and modulus of elasticity, the results of MLR and PLSR are not ideal.(4) Using discriminant analysis and Support Vector Machine (SVM) method to establish qualitative detection model of ripeness of watermelon based on the factors of vibration characteristic parameters and m. The results indicate that:the results of DA are similar to SVM, both the correct recognition rate is 88.2%. But the correct identification rates of ripe and over-ripe watermelons are small.(5) Using MLR method to to establish the quantitative regression model of hollow volumes based on the factors of vibration characteristic parameters and m, and establishing the qualitative detection model of watermelon hollow based on the DA method. The results showed that:1) The results of MLR method for the hollow volume are good, rc is 0.98, RMSEC is 41.1 mL, rp is 0.84, RMSEP is 35.9 mL.2) The correct classification rate of calibration set and prediction set of discriminant analysis method are 100%. |
PDF Full Text Request