Online System For Nondestructive Detection Of Pear Firmness Based On Acoustic Vibration

Firmness is an important quality indicator,and it is closely related to maturity and taste.Accurately measuring the fruit firmness is important for the determination of harvesting period,preservation after harvesting,product grading,variety breeding and the design of relevant fruit processing machinery.Traditional vibration detection sensors need to be attached to the surface of the object to be detected for signal acquisition,which is difficult to be applied to fast and real-time nondestructive detection of fruit quality.The laser doppler vibrometry(LDV)technology is a noncontact vibration detection method with high precision and fast response,and it has the potential for online detection of fruit quality.This paper applies laser Doppler vibrometry technology to the study of acoustical online nondestructive detection of pear firmness.The main research contents and conclusions are:(1)An online acoustic nondestructive detection system of pear firmness was developed and the system parameters were optimized based on the one-way ANOVA.The system is composed of an excitation unit,a detection unit,a delivery unit and a control unit.And optimal parameters of the excitation unit include the sound source aperture is 30 mm,the distance between the sound source and the measured object is 20 mm,the sound source output power is 40 w,the sweep:mode is linear and the sweep rate is 1000 Hz/s;The distance between the laser Doppler vibrometry and the measured object is 0.5 m,which is the optimal parameter of the detection unit.Based on the optimization of system parameters,the repeatability of measurement results was analyzed and it was found that the Differentiation index(D)was less than 10,indicating that the system has good stability and can be used to detect the pear firmness.(2)Taking Cuiguan pear as the test object,the vibration measurement was carried out,and the vibration spectrum was analyzed.The parameters such as resonance peak center frequency f,resonance peak width w,resonance peak area S,and peak resonance peak A were extracted.The reproducibility of vibration characteristic parameters was analyzed using the coefficient of variation of three measurements.The experimental results show that the coefficient of variation of all vibration characteristic parameters were less than 10%.And these four vibration characteristic parameters all have good repeatability,and can be used for the modeling of pear firmness.(3)Using a variety of chemometric methods,qualitative classification and quantitative prediction models of pear firmness were established.In the qualitative classification model of pear firmness established by discriminant analysis,K nearest neighbor classification algorithm,and back-propagation(BP)neural network,the BP neural network model has the highest discriminant accuracy.The accuracy of the correction set is 94.74%,and the accuracy of the forecasting set is 89.47%.Pear firmness quantitative prediction models were established based on multiple linear regression,principal component regression,partial least squares regression and BP neural network respectively.And the nonlinear model established by BP neural network method has the best prediction performance and is superior to the linear model established by other methods.And rp was 0.8726,RMSEP was 0.3420 N/mm.