Nondestructive Testing Technology Of Apple Moldy Heart Disease Based On Density And Near Infrared Spectrum

Apple moldy heart disease is one of common diseases in the process of apple growth,which occurs in the early stage,mature stage and storage period of fruit trees.It will cause fruit dropping,fruit quality degradation and fruit decay,and even lead to adjacent healthy fruit lesions and a series of phenomenon,which not only affect the safety of food,but also affect the apple’s market value.At present,apple’s external quality grading detection technology is relatively mature and widely used,but its internal quality detection technology is relatively poor.In order to realize the non-destructive detection of apple moldy heart disease,this study proposed a multivariate information detection method combining apple spectral characteristics with density characteristics.The main research contents are as follows:(1)The spectra was collected using MPA near infrared spectrometer.According to the characteristics of apple,the equatorial regions of apple samples of the spectrum were scanned every 120° time.To effectively eliminate the errors caused by different positions of spectrum acquisition,the average of three spectra was taken as the original spectrum of apple samples,making the prediction effects of the model more accurate.Canon camera 800D was used for image acquisition.After taking photos,the apple was rotated 90°and then photographed.Each apple was photographed for 4 times.(2)Convolution smoothing and mean centralization methods are used to preprocess the original spectral data.The above two preprocessing methods were used to preprocess the spectral data of apple samples ranging from 4000 to 12500 cm-1.The preprocessed spectral data were analyzed for principal component analysis,and the analyzed data were compared with Fisher discriminant model.It has been found that the accuracy of spectral data classification after convolution smoothing preprocessing is higher,so convolution smoothing preprocessing is the best preprocessing method.(3)By using the combination of Otsu and K-means clustering algorithm,the target area of sample apple was extracted accurately and its volume was calculated.The relationship model between the volume of the digital image algorithm and the real volume was established,and the real volume of the apple sample was calculated by the model.Finally,the method was verified by some samples,and the results showed that the error of apple volume was 6.52%,and the more accurate calculation of the apple volume was realized.Calculate the apple density using volume and mass according to the formula.(4)Combining the spectral and density information,SVM model,PSO-SVM model and BP neural network model were established respectively,and training set training model was used to test the accuracy of the model.The results showed that when the density+spectral information was used as input and PSO-SVM was used as training model,the accuracy was the highest(93.33%).Its prediction ability was better than other models,which can provide scientific guidance for the subsequent detection of apple moldy heart disease.Based on the near infrared spectrum and density characteristics,this paper has studied the detection method of apple moldy heart disease,and has made some new explorations on the nondestructive testing of apple,and the accuracy can basically meet the requirements,which provides theoretical support for the design and development of portable measuring instrument for apple moldy heart disease disease in the future.