Nondestructive Detecton Of Apple Internal Diseases Based On Electrical Characteristics

Non-destructive detection of Apple's internal disease is a major problem that puzzles apple's postharvest storage and the safety of the store and the economic impact of apple.In this study,the results of the 'Qin Guan' suspected of water heart disease,'Fuji' suspected mold heart disease and suspected internal browning fruit in the process of storage as the material,the same number of good fruit as a control,By using the method of the readings of 11 electrical characteristics in 100 Hz ~ 3.98 MHz to obtain the two types of fruit electrical parameters of the frequency spectrum.Then cut test to accurately distinguish sound and diseased fruit and determine the physical and chemical properties of the fruit.The principal component of the apple was extracted by the principal component analysis,and the three kinds of classification models were used to judge the accuracy of the non-destructive detection of the three diseases.The relationship between the physical and chemical properties and the incidence of disease was analyzed.In order to provide the theoretical basis and technical support for the rapid non-destructive detection of apple internal diseases.The main results are as follows:(1)Through the determination of electrical characteristics of watercore apple and sound fruit,it was found that the value of complex impedance(Z),dielectric loss coefficient(D),equivalent series inductance(Ls),conductance(G),complex impedance angle(deg),equivalent series capacitor(Cs),parallel inductance(Lp),parallel resistance(Rp)and parallel capacitor(Cp),equivalent series resistance(Rs),susceptance(B)and relative permittivity(?')and the loss factor(?")the same trend observed in 100 Hz ~ 3.98 MHz frequency range of each frequency point.In the low frequency region 1 00 Hz ~ 10000 Hz,the indexes D,deg,Cs,the Cp,G,?',?' of watercore apples were significantly higher than the sound fruits.Through principal component analysis to extract 15 principal components value from 143 characteristic electrical parameters were determined directly.Using Fisher discriminantanalysis,multilayer perceptron(MLP)neural network model and radial basis function(RBF)neural network model to identify sound fruit and watercore apples,which identify using the MLP neural network model can achieve the highest rate of 95.4%.With the loss factor as a variable,the identification of the three models can reach 100% at one or several frequencies.The density,hardness and soluble solids of watercore fruit were significantly higher than that of good fruit titratable acid content was significantly lower than that of good fruit(p < 0.05).(2)Though the study of electrical characteristics of internal browning fruit and sound fruit,we found that changes of watercore apple and sound fruit are similarity.In the low frequency region of 100 ~ 39800 Hz,deg,D,Cs,Cp,?',?' indexes of disease fruit were significantly higher than the sound fruit.Through principal component analysis,14 principal components are extracted,using three modeling methods to identify the browning fruit disease,the identification of sound fruit and browning fruit by using MLP neural network model is the most accurate,and the highest rate can reach 100%.The hardness and Vc content of browning fruit were significantly lower than sound fruit(p < 0.05).(3)The study on the electrical characteristics of the heart disease and sound fruit,found that 13 electrical indicators were same as the frequency trend,in the low frequency the D value of modly fruit slightly higher than sound fruit and the difference is not obvious.The deg value of the diseased fruit in the 100-3980 Hz region is slightly larger than the sound fruit while the high-frequency 3980-3980000 Hz area is less than the good fruit,but the difference is not obvious.The value of Cp,Cs and ?'' in the low-frequency area of the diseased fruit and the sound fruit is significantly less than the sound fruit,which tends to be consistent in the high frequency area.For ?',the sound fruit value is significantly higher than the disease fruit.Through the main component analysis,23 main components were extracted,and three kinds of modeling methods were used to identify the moldy core disease.The MLP neural network model was found to be the best model to recognize moldy core and sound fruit.The density and soluble solids of the moldy core fruit were significantly lower than that of sound fruit(p <0.05)...