Nondestructive Evaluation Of Pear Texture Based On Laser Doppler Vibrometry

The pear industry is the third largest fruit industry in China. Along with science and technology advancement and the improvement of people’s living standards, the demands for high fruit quality become more and more critical. Fruit texture, as a key factor for fruit texture evaluation, reflects the rheological and structural properties of fruit. Pear, which is a kind of respiration climacteric fruit, continues to soften in the maturation process. Accurate detection of pear texture is important for the optimum harvest date determination, postharvest storage, optimum edible period evaluation, product grading, and fruit processing machinery design. The sensory evaluation for fruit texture has some drawbacks, such as being subjective, diversity among individuals, destructive detection and high human cost. Traditional instrument measurement also has defects of destructive detection, lower detection efficiency and huge waste. Therefore, it is essential to find a nondestructive detection method for fruit texture evaluation.The acoustic vibration technique is the most commonly used nondestructive detection method for the evaluation of fruit texture among the existing researches. Contact sensors are attached directly to the surface of fruit to detect the vibration. However, the contact sensor may impair the free vibration of fruit and damage the fruit surface, and it can not be used on many occasions. Therefore, the scope of application of contact sensors is limited. The laser Doppler vibrometry (LDV) technology, as a non-contact measurement, has the advantages of high accuracy, quick dynamic response, wide measure range, not being affecting by the ambient noise, and high sensitivity to the vibration displacement. These features meet the demands of fruit vibration measurement.Taking the pear as the research object, the study used the mechanical vibration, sensing technology, signal analysis and processing technology, and chemometrics together for the research on nondestructive detection of pear texture based on the LDV technology. First, a testing platform based on the LDV technology was established for fruit vibration characteristics measurement. Secondly, pear dynamic characteristics under the swept sine vibration excitation were analyzed by the finite element (FE) harmonic response analysis, which provided the theoretical basis for pear dynamic characteristics measurement by the LDV method. Thirdly, the feasibilities of nondestructive detection of pear texture by the LDV method with the frequency sweeping mode and impulse response mode were investigated. Different modeling methods were compared. The proposed approach provides a way for rapid detection of pear quality to meet the requirement of on-line detection. The aim of the study is to prove the feasibility of the LDV method for nondestructive evaluation of pear texture and establish a corresponding research system. It will provide the basis for the research and development of pear texture detection equipment and detection line.The main results and conclusions were listed as follows:(1) A testing platform for fruit vibration characteristics based on the LDV technology was established, and the test parameters under the frequency sweeping mode were optimized. The platform includes the vibration control part, vibration signal acquisition part, and signal processing part. The results showed that:1) The frequency-response curves at the same detection point have good repeatability under all tested frequency sweeping rates (regardless of the linear or logarithmic frequency sweeping mode) and acceleration amplitudes. However, the frequency-response curves at different detection points have poor repeatability for each laying style.2) The second resonance frequency (f2) has good repeatability no matter what test condition is taken. P800, P1200, and P1600 show relatively good repeatability.3) The better combination of test parameters for better repeatability were obtained by the three factors and three levels orthogonal experiment, which are as follows:the frequency sweeping rate of 1400 Hz/min with the linear frequency sweeping mode, acceleration amplitude of 1g (g=9.8 m/s2), and laying style that the pear is placed with its stem upward.(2) A 3-dimensional model of the pear was then established by a machine vision-based modeling system. Then, the pear dynamic characteristics were analyzed by the FE harmonic response analysis. The results showed that:1) The resonance frequencies of the FE models obtained by the model updating technique agreed well with the measured frequencies, indicating that the FE harmonic response analysis to investigate the pear dynamic characteristics under forced vibration is feasible.2) The mode shapes corresponding to the second to fifth resonance frequency (f2～f5) showed simultaneous extension and contraction in two mutually perpendicular directions, but no torsional deformation were observed. The maximum deformation occurred at the end of the stem and calyx, so it would be better to measure these modes at the end of the stem or calyx by the LDV method.3) The resonance frequencies increased with increasing Young’s modulus and decreased with increasing density, Poisson’s ratio, mass and height/diameter (H/D) ratio, and mass and Poisson’s ratio had less effect.(3) The time-course changes in the vibration characteristics under the swept sine vibration excitation and texture of pears during storage were analyzed, and the results showed that 1) Among the vibration parameters, elasticity indices (El) declined gradually over time; phase shifts increased gradually over time; but the amplitudes at resonance frequencies did not show obvious variation trend during storage.2) All the texture indices, including Magness-Taylor (MT) firmness, flesh firmness (FF), and stiffness (Stif), declined over time during storage. Moreover, Stif varied more smoothly, and was more sensitive to the storage time than MT and FF. The correlations between the texture indices and the vibration parameters of 3 pear varieties were analyzed, and the results showed that 1) There were no obvious difference and regularity in correlation coefficients of the vibration parameters with peeled and unpeeled texture indices. 2) Stif had the highest correlation coefficient with the vibration parameters among the texture indices.3) EI had the highest correlation coefficient with Stif among the vibration parameters, followed by phase shifts. However, there was no correlation between the amplitudes at resonance frequencies and Stif. The LDV method and the traditional puncture test were compared. The results showed that the fruit texture can be detected by the LDV method, and the LDV method is superior to the puncture test in terms of repeatability and sensitivity.(4) Quantitative models for Stif were established. Based on the removal of spectra and concentration outliers, the Stif prediction model was obtained by the unary linear regression (ULR) and stepwise multiple linear regression (SMLR) methods, and a modified modeling method based on fruit shape was proposed. The results showed that:1) The SMLR model was superior to the ULR model.2) The performance of the prediction model was improved after the introduction of fruit shape index, H/D ratio. After the introduction of H/D ratio to the SMLR model, the value of correlation coefficient for the calibration sample set (rc) increased from 0.850 to 0.862, root mean square error of calibration (RMSEC) decreased from 0.811 N/mm to 0.781 N/mm, the value of correlation coefficient for the prediction sample set (rp) increased from 0.760 to 0.780, and root mean square error of prediction (RMSEP) decreased from 0.954 N/mm to 0.914 N/mm. The results indicated that nondestructive detection of pear texture by the LDV method with the frequency sweeping mode is feasible. In addition, the feasibility and universality of the LDV method for texture evaluation of different pear varieties and pears in different years were investigated. The results showed that the LDV method shows a good universality for pears of different varieties and years.(5) The feasibility of the impulse response method using a laser Doppler vibrometer for pear texture and freshness evaluation was investigated. The results showed that:1) There was no damage to the pear when the impulse amplitude of the half-sine impulse signal was less than 4g. 2) The 15 features in time domain and 8 features in frequency domain extracted from the time and frequency domain signals have good repeatability.3) Based on the removal of spectra and concenteration outliers, the SMLR method, back-propagation neural network (BPNN), and principal component analysis (PCA)-BPNN were used for quantitative analysis of pear texture. The results showed that the PCA-BPNN showed the best modeling result among the used modeling methods. The rp of MT, FF, and Stif were 0.625,0.840, and 0.577, respectively. RMSEP of MT, FF, and Stif were 1.892 N,0.959 N, and 0.838 N/mm, respectively, indicating that the impulse response method using a laser Doppler vibrometer for pear texture evaluation is feasible.4) Based on the removal of spectra and concentration outliers, the Fisher discriminant analysis (FDA), BPNN, and PCA-BPNN were used for pear freshness discrimination. PCA-BPNN showed the best discriminant result, by which the discriminant accuracy for the calibration sample set was 89.0%, and that for the prediction sample set was 83.3%. The results indicated that the impulse response method using a laser Doppler vibrometer for pear evaluation is feasible.