Nondestructive measurement of pear texture by acoustic vibration method

• We studied the relations between different texture indices and vibration parameters.
• Stiffness was found to be more suitable for evaluating pear quality.
• A new modeling approach for predicting stiffness was proposed.
• Three vibration parameters for predicting stiffness were extracted by SMLR method.
• The performance of the prediction model was improved by introducing shape index.

Texture is a key attribute for the assessment of pear quality, and a nondestructive texture detection method was investigated. Each pear fruit was excited by a swept sine wave signal (xin), and the response signal from the top of the pear (xout) was detected by a laser Doppler vibrometer (LDV). The vibration spectrum was acquired after a fast Fourier transform was applied to the xin and xout data. Six vibration parameters, including the second resonance (f2), the amplitude at f2 (A2), and the phase shifts at 400, 800, 1200 and 1600 Hz (P400, P800, P1200 and P1600) were extracted from the vibration spectrum, and the elasticity index (EI) was determined by the formula EI=f22m2/3. The fruit texture was then measured by a puncture test. Three texture indices were extracted from the force–deformation curve, in which the stiffness (Stif) was found to be more suitable for representing fruit quality. The multiple linear regression (MLR) method was applied to evaluate the importance of each vibration parameter for predicting Stif, and the following order of importance was found: EI, f2, P400, P1600, P800, P1200, and A2. A texture prediction model was built by the stepwise multiple linear regression (SMLR) method and modified through the introduction of the pear shape index (SI). The performance of the prediction model was improved after modification; the value of the correlation coefficient for the calibration and validation sample sets (rc and rp) increased by 0.4% and 2.1%, respectively, while the root mean square errors of calibration and prediction (RMSEC and RMSEP) decreased by 0.6% and 3.3%, respectively. Highly significant results (P < 0.01) for both the initial and modified prediction models proved that the evaluation of pear texture by a combination of the LDV method and the proposed approach was feasible.