Use of Biospeckle Imaging to Measure Mechanical Properties of Apples in the Non-invasive Manner

Introduction The texture of fresh fruit is determined by the structural and mechanical properties of tissue. It depends on climate, maturity, variety and postharvest condition. During ripening, due to loss of turgor, degradation of starch and cell walls, the flesh of apple softens. The relationship between fruit quality and its physiological changes has been widely investigated. Using techniques according to the principles of force-deformation, impact, and vibration tests, texture of fruit and its mechanical properties can be associated, conventionally. In analyzing the vitality of biomaterials; a non-invasive technique based on the optical phenomenon is the Biospeckle method which occurs when the surface of the sample is illuminated by laser light. It seems that because of the fact that the laser light can penetrate tissue, it is possible to obtain information about the texture and cell condition from tissue under the skin. This means that, there would be a chance to detect and monitor the variation of cells and try to make a model to predict mechanical properties. Therefore, the overall objective of this study was to develop prediction models based on biospeckle imaging to predict mechanical properties of ripe Golden Delicious apples.   Materials and Methods The 400 fresh and intact 'Golden Delicious' apples were harvested and were prepared for mechanical tests and biospeckle imaging. Biospeckle imaging was carried out first, followed by compression and creep test and then penetration test. During imaging, to avoid environmental reflections, the process was carried out in a dark and closed chamber. Biospeckle activity was saved as a video (AVI format) in a computer for analyzing. The THSP method was used to analyze biospeckle activity in samples. The indices which have been used for analyzing biospeckle images are divided into 3 statistical features and 4 textural features. Apples were cut in half. One of the halves was used for cylindrical sample extraction for uniaxial compression and creep tests and another was used for penetration test. From compression tests the tangent modulus of elasticity, stress and strain of bio-yield and failure energy for toughness calculation were determined. The creep behavior was obtained by fitting the Burger's model to the experimental data. In penetration test, a stainless steel probe with a hemispherical tip was used for peeled and unpeeled samples. For each sample maximum penetration force and energy were obtained. Prediction of mechanical property was carried out using adaptive neuro-fuzzy inference system (ANFIS). To reduce the dimension of the input vector the PCA was used. Four significant adjustments were made in the structure of ANFIS in order to find the best models. The models were evaluated using RMSECV, RMSEP, MBEC, MBEP, RC, and RP.   Results and Discussion Models for modulus of elasticity prediction have Rp=0.821, 0778, 0.791, 0.880, and 0.843 for 4 compression rate and secant modulus, respectively. Clearly, the results from this research are encouraging, indicating the potential of using speckle imaging system for predicting apple fruit mechanical properties. Comparing to the all texture analysis techniques, Wavelet and GLRLM provided good results for most properties leading to select them as the best techniques for analysis of biospeckle images because of their consistency in prediction performance. Prediction model for break strain has the highest Rp (Rp=0.920) followed by the retarded time (Rp=0.890), retarded viscosity (Rp=0.886) and maximum penetration force in unpeeled case (Rp=0.883). A lower correlation (Rp = 0.728) was observed for initial viscosity. Conclusions The described optical method based on biospeckle represents an innovative and reliable method for rapid and non-invasive detection of mechanical properties. The results of the evaluation showed that, as time passes, fresh apples due to the loss of water in both the elasticity and the biospeckle activity were dropped. Biospeckle imaging can accurately predict mechanical properties. The average accuracy of best prediction of mechanical properties models was R2=0.899. The present results can provide the basis of future development of in-line quality monitoring during apple quality control.