| Crispness is one of the important indexes to reflect the internal quality of Korla pear.At present,sensory evaluation is the main detection method of crispness,but sensory evaluators are prone to fatigue and low evaluation efficiency,and the evaluation results are easily affected by subjective factors.Therefore,it is urgent to find an objective,efficient and accurate method to evaluate the crisp degree of Korla pear,so as to realize the internal quality detection of Korla pear closer to the taste of consumers.In this paper,a texture analyzer is used to perform puncture tests on the Korla pear pulp to obtain mechanical signals,and the AED low-frequency and recording high-frequency acquisition methods are used to obtain the acoustic signals of the Korla pear pulp puncture.The peak method is used to obtain the characteristics from the mechanical signal and the acoustic signal.1/3 octave method is used to extract frequency domain characteristic parameters from high frequency sound signal.Through the significant difference analysis and the correlation analysis with the Korla pear sensory crispness evaluation results,the characteristic parameters that can characterize Korla pear crispness are extracted.The parameters were analyzed by autocorrelation and reduced by principal component analysis.Finally,based on multiple linear regression and BP neural network,the crisp degree detection model of Korla pear was constructed to realize the objective evaluation of crisp degree of Korla pear.The main results of this research are as follows:(1)A texture instrument was used to collect the mechanical signal of the puncture of the Korla pear pulp.The low frequency of 500Hz was acquired by AED and the high frequency of 44100Hz was acquired by recorder respectively.The sound signal collected by recorder is denoised by spectral subtraction.The frequency domain signal is obtained by FFT of the de-noising signal.(2)15 mechanical parameters are extracted from the mechanical signal by the peak method.The young's modulus,low strain stiffness,force ratio,average force,force work,number of force peaks,force linear distance and terminal force showed a significant downward trend with the increase of storage period,and there were significant differences in different storage periods.The first force drop and force difference showed an upward trend with the increase of storage period,which can be used as the mechanical characteristic parameters to evaluate the crispness of Korla pear.(3)The number of sound peaks,sound power,maximum sound pressure level,average sound pressure level,sound linear distance and the sound pressure level at the first fracture of the 6 extracted AED acoustic parameters.When the Korla pear crispness decreases with the increase in the storage period of the Korla pear Overall,there was a significant downward trend;The maximum sound amplitude,peak number,standard deviation,length of sound curve,amplitude difference and average peak decreased significantly with the increase of storage period of Korla pears.All of them can better reflect the changes of Korla pear pulp crispness,and these parameters can be used as characteristic parameters for evaluating Korla pear crispness.The 1/3 octave method is used to extract 32 power spectrum characteristic parameters from the high-frequency collected sound and audio domain signals.(4)The autocorrelation analysis of the selected force acoustic parameters shows that there is a high correlation among the force and sound parameters.Therefore,the dimension reduction is carried out by principal component analysis,and three principal components are extracted to explain the mechanics and AED acoustic characteristics parameters 83.87% of the information;Four principal components to explain the mechanics and high-frequency acquisition of the acoustic time domain feature parameters of 83.30% of the information;Nine principal components to explain the mechanics and high-frequency acquisition of the acoustic frequency domain feature parameters of 82.07% of the information.(5)The crisp degree evaluation model of Korla pear was established by multiple linear regression based on principal component analysis after dimension reduction.All three models can accurately evaluate the Korla pear crispness(r>0.800).The correlation coefficient of the model prediction results constructed by the principal components of mechanics and AED acoustic parameters and the sensory crispness score was 0.850,and the RMSE was 0.521;the prediction results of the model constructed by the principal components of the mechanical and high-frequency acquisition acoustic time domain parameters are consistent with the sensory crispness score correlation coefficient r was 0.830,and the RMSE was 0.585;the main component prediction results of the acoustic frequency domain parameters using mechanical and high frequency acquisition and the sensory crispness score correlation coefficient r was 0.803,and the RMSE was 0.589.(6)The BP neural network was used to construct a crispness assessment model based on principal components after dimension reduction,The model model constructed by the principal components of mechanical parameters and AED acoustic parameters predicts that the correlation between Korla pear crispness and sensory crispness r was as high as 0.872,and the mean square error MSE was 0.582;The model is constructed by principal components of mechanical and high-frequency acquisition acoustic time-domain predicts that the correlation r between Korla pear crispness and sensory fragility is as high as0.873,and the mean square error MSE was 0.645;The model constructed by mechanical and high-frequency acquisition of acoustic frequency domain characteristic principal components predicts the correlation r between Korla pear crispness and sensory crispness up to 0.853,MSE was 0.795.In this study,the mechanical and acoustic synchronous testing method proposed can realize the accurate detection of the crisp degree of Korla pear pulp.Extracting characteristic parameters from the mechanical and acoustic signals of Korla pear pulp puncture to construct Korla pear crispness detection model,which is of great significance to realize the accurate grading of Korla pear internal quality close to the taste of consumers. |
PDF Full Text Request