Nondestructive Methods And Instruments For On-site Inspection Of Peach Firmness Based On Mechanical And Acoustic Vibration Properties

China is a major producer and grower of peaches,yet the peach industry has been facing a problem of prioritizing quantity over quality in recent years.Fruit firmness is a critical indicator of peach quality.In-situ firmness detection of on-tree peaches can aid in improving peach quality to meet market demand.The mechanical and acoustic vibration properties of fruit are closely related to its firmness.In orchards,peaches are usually picked by sampling and their rupture force is measured using a handheld penetrometer to characterize firmness.But its measurement results are influenced by the loading force,loading speed,and measurement point of the operator.As a result,the detection accuracy of the penetrometer is not high.In-situ detection of peach firmness can provide useful data for breeding high-quality varieties,without disturbing fruit growth.The acoustic vibration detection method has the potential to achieve in-situ detection of peach firmness.However,it is limited by the unclear mechanism of in-situ acoustic vibration detection,the lack of non-contact instruments,and the absence of effective methods for analyzing acoustic vibration response and constructing peach firmness detection models.To solve these problems,this paper took peaches as the research object and investigated on-site detection methods and instruments based on the mechanical and acoustic vibration properties of peaches.This study aimed to meet the demand for sampling and in-situ detection of peach firmness in orchards and to provide technical support for improving peach quality.The main research contents,results,and conclusions are as follows:(1)The changing pattern of peach firmness during growth was investigated and an operator-independent detection method of peach firmness was established.Portable instruments for detecting peach firmness based on force perception were developed.The results showed that both the rupture force and the initial slope extracted from the force-displacement curve exhibited an obvious decreasing trend in the late growth stage of peaches.To characterize the change of peach firmness during growth,the initial slope was selected since it reflected the fruit elasticity and could be measured nondestructively.To avoid the influence of the operator’s loading force and speed on measurements,a manual loading instrument for peach firmness detection based on force perception was developed.A limit structure was designed inside the instrument to provide a constant loading force for the probe.And the mapping relationship between the prototype reading and peach firmness was established based on the force analysis.The repeatability(CV_p)and reproducibility(CV_d)indexes of the manual loading prototype were less than 0.5%.The prototype reading showed a significant positive correlation with the initial slope(R~2=0.8971,p<0.01),which corroborated the results of the force analysis.To further enhance the automation of the instrument,an automatic loading prototype was developed.The results showed that the standard deviations(RSDs)of the measured values by the same operator using the automatic loading prototype were less than 0.61%.And there was no significant difference between the measured values of different operators(p>0.05).Additionally,the automatic loading prototype outperformed the commercial handheld penetrometer for measuring peach firmness.Theand RMSEP of the regression model between the prototype reading and the peach firmness were 0.923 and 1.613 N/mm,respectively.The visualization results indicated that the firmness of different parts varied,confirming the need to measure the firmness of the whole peach.(2)To analyze the relationship between the acoustic vibration response and the firmness of the whole peach,a peach vibration model was established.This study investigated the acoustic vibration response mechanism of on-tree peaches and built a non-contact instrument for measuring their acoustic vibration response.The results showed that both the second and third natural frequencies of the peach were positively correlated with the peach firmness(r=0.978,p<0.01).The acoustic vibration response of the peach was affected by external constraints,with larger constraint areas resulting in higher natural frequencies.The excitation force did not affect the position of the resonance peak.Random interference caused burrs in the peach power spectra and drowned out the resonance peaks,making identification difficult.Therefore,noise reduction for the measured acoustic vibration response was necessary.The position of the excitation and measurement points affected the amplitude of the power spectra,but had little effect on the location of the resonance peaks.The variation of the second and third resonance frequencies was between 2.61%to 6.95%.Based on the above findings,a non-contact instrument was developed to measure the acoustic vibration response of peaches on trees.The gas pressure,excitation time,and excitation distance of the instrument were optimized to 300 k Pa,200 ms,and<20 cm based on the magnitude of the excitation force and the signal-to-noise ratio of the acoustic vibration response.At the same time,the non-contact laser Doppler vibrometer was used for nondestructively measuring the acoustic vibration response.(3)Furthermore,this study proposed a method for analyzing measured acoustic vibration response and established a deep learning model with multi-scale perceptual fields for detecting peach firmness.The time-frequency properties of the acoustic vibration response of an on-tree peach were analyzed by the wavelet transform method.And the wavelet threshold denoising method was established for noise reduction.The optimal wavelet threshold denoising parameters were determined by comparing the root mean square error(RMSE),signal-to-noise ratio(SNR),and smoothness index(r).The results showed that the db6 wavelet function with 7 decomposition levels and Heuristic sure threshold combined with the soft thresholding function achieved the best denoising results(RMSE=0.016,SNR=8.137,and r=0.739).On this basis,the power spectral density(PSD)of the acoustic vibration response was obtained by the periodogram method,autocorrelation method,Welch method,and AR model method.The results showed that the smoothness and clarity of the PSD obtained with the AR model method were the best.Based on the Akaike information criterion(AIC),the final prediction error(FPE)criterion,and the minimum description length(MDL)criterion,the optimal order of the AR model was determined to be 15.Furthermore,a novel one-dimensional convolutional neural network(CNN_m)with multi-scale perceptual fields was constructed by fusing the Inception module.It was compared with the peach firmness detection model based on the partial least squares regression(PLSR)and support vector regression(SVR)methods by the RMSE,coefficient of determination(R~2),and residual prediction deviation(RPD).The results indicated that the SVR model based on the whole PSD outperformed the PLSR model.And CNN_m predicted peach firmness more accurately than other models.The,RMSEP andof CNN_m were 0.813,2.501 N/mm,and 2.334,which outperformed the SVR model by 9.0%,14.2%,and 16.5%,respectively.