| Apple core rot is a serious internal disease of fruits.The appearances of fruits have no obvious disease characteristics,which seriously restricts the development of apple industry in China.Therefore,it is of great significance to establish a fast and reliable apple core rot non-destructive testing model in practical production.In this paper,fuji apples were taken as the research object.The nutritional quality,physiological quality,sensory quality,texture quality,volatile substance,aging indices and enzymatic defense system of dry core rot apples(moderate)and wet core rot apples with different degrees were measured to explore the quality difference between them.In additon,based on the electronic nose,a discriminant model of healthy and core rot apples was established and the discriminant effect of different models was compared.Based on diffuse transmission spectrum and diffuse reflectance spectrum,the discrimination model of core rot apples with different degrees was established.Moreover,the best detection model of apple core rot was obtained by comparing the different spectrum acquisition methods,preprocessing methods,modeling methods and modeling effects of different modeling bands of near-infrared spectrum.The test results are as follows:1.Apples with different pathological types and different grades of wet core rot had different quality and physiological and biochemical reactions.For different types of lesions,compared with the wet core rot apples,the density of dry core rot apples of the same disease degree was low,which promoted the decrease of fruit firmness and brittleness,but the kind and content of volatile substance(aldehydes,esters and alcohols)were higher.In addition,the respiration intensity of dry core rot apples was lower.MDA and relative conductivity were suppressed.Polyphenol oxidase(PPO)and peroxidase(POD)activities were kept at lower levels.Total phenol content and phenylalanine ammonia-lyase(PAL)activity levels were higher.These all delayed the senescence and browning of fruits,leading to the appearance of different pathological types.For different grades of wet core rot apples,the higher the degree of pathological changes,the lower the density of apple fruit,which made the fruit's titratable acid content,texture quality(peel strength,peel brittleness,firmness,brittleness and chewability)and the content of esters were gradually reduced.But the respiratory intensity and the content of alcohols and aromatic hydrocarbons weregradually increased.In addition,the MDA content and relative conductivity can be promoted with the gradually increase in the degree of lesions,and a higher total phenol content and a higher activity of PPO,CAT,POD and PAL could be maintained.2.Based on SIMCA software,principal component analysis,cluster analysis and orthogonal partial least squares-discriminant analysis were used to analyze the electronic nose information of different grades of core rot apples.The results showed that W5 C,W1C and W3 C were the characteristic sensors of healthy fruit and W1 S,W2S,W1 W and W5 S were the characteristic sensors of severely diseased fruit.Based on SPSS 23.0 software,the fisher discriminant function,muhilayer perceptron neural network(MLPNN)and radial basis function neural network(RBFNN)discriminant models of apple core rot were established.The results showed that the model of electronic nose combined with MLPNN had the best discrimination effect,and the overall prediction accuracy for the modeling and validation were 88.61% and 88.46%,respectively.The model of the electronic nose combined with RBFNN was secondary,which the overall prediction accuracy for the modeling and validation were 93.50%and 80.95%,respectively.The model of the electronic nose combined with fisher discriminant function had the worst discrimination effect,and the overall prediction accuracy for the modeling and validation were 91.50% and 79.27%,respectively.In addition,the three discriminant models had good discriminant effect on healthy and severe fruit,and the discriminant effect on healthy and mild fruit was not ideal,which needs to be further optimized in the future research.3.Based on the diffuse transmission spectrum,three kinds of spectral acquisition methods,five kinds of spectral pretreatment methods and four kinds of modeling methods were used to establish the nondestructive testing model of apple core rot.The results of two-point transverse modeling showed that the model established by unit normalization combined with PLSR was the best,with a discrimination rate of85.29%.The results of four-point transverse modeling showed that the model established by unit normalization combined with PLSR was the best,with a discrimination rate of 86.27%.The results of four-point vertical modeling showed that the model established by unit normalization combined with MLR was the best,with a discrimination rate of 83.33%.In conclusion,the four-point transverse spectrum acquisition method,unit vector normalization preprocessing method and PLSR modeling method were more suitable for diffuse transmission spectrum to establishthe nondestructive testing model of apple core rot and the discrimination rate was86.27%.Then the four-point transverse diffuse transmission spectrum was modeled in different bands.The modeling results showed that the four modeling methods in the band of 710 to 950 nm had good discrimination effect,and the discrimination rate of PLSR was as high as 87.25%.4.Based on diffuse reflectance spectroscopy,5 spectral preprocessing methods and 4 modeling methods were used to establish a nondestructive testing model of apple core rot.The results showed that the first derivative combined with PLSR had the best effect,with the discrimination rate of 83.74%.The effects of the first derivative combined with MLR and SVR were secondary,and the discrimination rates were 82.11% and 81.30%,respectively.The discrimination rate,between 60.16% and67.48%,of PCR model was relatively low.Therefore,PCR was not suitable for establishing apple core rot model by diffuse reflectance spectroscopy.The results of modeling in different wavebands showed that the model of the second band(740 nm~ 1386 nm)was the best,and the discrimination rates of MLR,PLSR and SVR were82.93%,80.49% and 79.67%,respectively.Comparing the results of diffuse transmission spectrum and diffuse reflection spectrum modeling,no matter the original model or the optimal pretreatment model,the diffuse transmission spectrum was more suitable for the establishment of non-destructive testing model of apple core rot. |
PDF Full Text Request