Research And Evaluation On Watercore Discrimination And Shelf-life Prediction Of Apple Based On Non-destructive Detection

The water-core non-destructive detection of apple and the prediction of fruit’s shelf life is a big problem which trouble postharvest industry and market liquidity of apples. In this study, hyperspectral imaging technology and dielectric properties’ detection were used to collect near infrared spectroscopy and measure dielectric parameters for water-core fruit and sound fruit of apples cv. Qinguan based on the compatible test system conditions. The characteristic bands of near-infrared spectroscopy and dielectric parameters which represent of water-core fruit and discriminant model for water-core can be screened and established by analyzing water-core fruit and sound fruit’s performance at different bands or dielectric parameters.In order to establish shelf-life prediction model based on quality index and dielectric properties of fruit, we measured apple’s dielectric properties and quality indices’ dynamic variation during storage at 20℃and 0℃ and found the threshold value of quality index during storage, then analysed the relation of indexs and dielectric properties. The shelf-life prediction model would offer a theory to determine the maximum shelf life and the best delivery time of apple.The results were shown as follows:(1) Four kinds of feature selection methods based on chi-square test, F classic test,SVM-RFE,decision tree and 3 kinds of kernel function of support vector machine(SVM) classifier included SVM-linear, SVM-poly, SVM-rbf were adopted to recognize watercore fruits according to the characteristics of hyperspectral of fruit. The accurate rate of watercore distinguish of 4 kinds of feature selection with 3 kinds of kernel function of SVM classifier at 1~200 wavebands was 48.6%~70.2%, 48.6%~72.0%, 33.3%~71.8% and 47.2%~70.8%, respectively; moreover,the accurate rate of watercore distinguish based on SVM-RFE reached the highest level of 72% at band of 126(1356.47nm) and the accurate rate of watercore distinguish based on SVM-rbf was more steady. Kullback–Leibler divergence and random forest were adopted to recognize watercore fruits according to dielectric properties of fruit. The selection method and random forest classifier for detection of apple watercore using dielectric properties. The accurate rate of water-core distinguish based on dielectric parameters was reached the highest level of 76.67% at the first 30 dimensions.(2) During storage, the changes of weight-lose rate and hardness of fruit with storage time had a goodness of fit and the equation reached a significant level. By comparing the difference of indexs(such as, respiration intensity, weight-lose rate, hardness) at 20℃and 0℃storage and then converted into aging rate ratio of apples cv. Qinguan. The results showed that aging rate ratio at 20℃ was more 3.83, 3.57 and 3.17 times, respectively, than that of 0℃. Finally, we chosed the weight-lose rate of 11.35% and hardness of 3.44kg/cm2 as the quality indicator threshold according to the standard of sensory evaluation and tasting which took fruit appearing wrinkled and not crisp as a standard, combined with reverse tracing method.(3)The correlation between weight-lose rate,hardness and complex impedance Z,series inductance Ls,parallel inductance Lp,susceptance B reached a significant level(P<0.01)under storage at 20℃with the frequency of 1k Hz and 0℃with the frequency of 1000k Hz.The relative error of predicting on shelf-life of apple which used the relation of 4 sensitive dielectric parameters and weight-lose rate was lower than the relation of 4 sensitive dielectric parameters and hardness.Among all the dielectric parameters,the relative error of series inductance Ls predicting on shelf-life of apple was lowest,which was just 1.11%at 20℃.Therefore,the fitting equation of Y失重率=-54.313lg Ls1000+141.56 and Ylg Ls=﹣0.003X+2.5885 as a reference to predict the shelf life of fruit under 20℃temperature storage.We found the shelf-life of apples cv.Qinguan was over 200d at 0℃through combining the actual storage time,which based on the quality indicator threshold.