Research On Non-destructive Testing Method And Equipment For Apple Ripeness Based On Multi-indicator Factors

The ripeness of the fruit is one of the key factors influencing the development of fruit ripeness and taste and flavor,and directly determines the commercial value of its fresh apples after harvest.Equally important for apples in storage,appropriate harvesting is an important technical step to ensure fruit quality and extend the post-harvest life of the fruit.This indicates that apple maturity is very important for both fresh apples and apples in storage.Therefore,in this paper,apples from the same commercial orchard in Fubeng County,Shaanxi Province,during the ripening period from 2019 to 2021 were used as samples to carry out research on apple ripeness evaluation methods and nondestructive testing of apple ripeness using visible near-infrared spectroscopy and embedded technology as tools,and to design and develop portable testing equipment for the practical application of rapid nondestructive testing of apple ripeness.The main conclusions of this thesis were as follows:(1)A multi-indicator factor method for apple ripening assessment based on radar charts was investigated.By measuring the physical and chemical indicators of apples during ripening,it was found that the soluble solids content and color value a*of apples would increase with the increase of apple ripening,and the rest of indicators would decrease.Correlation analysis showed that there was a high correlation between peel colors.Finally,apple soluble solids content,titratable acid,color value b*,average flesh firmness,and maximum peel force of peel breakdown were screened as important factors for evaluating apple ripeness.Based on the variation pattern of ripening factors in apple ripening process and the geometric characteristics of radar map,we explored the visual evaluation index of apple ripening(VRPI).By comparing with Streif index,SPI index,RPI index and IQI index,it was found that the visual evaluation index of apple ripening had a high correlation with all other indexes,especially IQI index,but this index was less influenced by skin color than IQI index,which could distinguish apples of each ripeness level more clearly than Streif index.(2)A series of models for predicting VRPI was developed based on visible near-infrared spectroscopy.The spectral models for predicting apple ripeness were developed based on full spectrum and feature wavelength using partial least squares algorithm,convolutional neural network,support vector machine,least squares support vector and random forest algorithms,and the correlation coefficients of the prediction models(R_p)ranged from 0.833 to 0.925 with root mean square errors()between 0.168 to 0.247.A decision-level data fusion study of the model results showed that theR_pof the fused models ranged from 0.871 to 0.913 and theranged from 0.184 to 0.213 for the same year samples.For the other years of the sample,the fused model was less affected by the predicted results of the single model.(3)The impact of sample year differences on the spectral model was investigated.The slope bias correction method can reduce the deviation in the prediction process of the model,but can not improve theR_pof the model.The degree of reduction for thereached18.83%～37.08%.The global model was the most beneficial in eliminating the effect of year on the spectral model.When the model containing the 2019 and 2020 samples predicted the2020 sample,the prediction accuracy improved fromR_p=0.848 and=0.561 toR_p=0.906 and=0267.For the unknown sample in 2021,theR_pof this model improved from 0.811 to 0.866;theturned from 0.478 to 0.508,and further used slope bias correction to reduce RMSEP to 0.331.This study provided a methodological reference for reducing the impact of year variation on apple ripeness visualization index detection models.(4)The portable apple ripeness testing device was designed and developed.The equipment mainly included five parts:micro-spectrometer,micro-control circuit board,rechargeable lithium battery,tungsten-halogen lamp and display screen,etc.The optimal integration time of the device was 6 ms.It took 2 s for the device to complete the detection of apple maturity.The precision of the model in this device was robust(R_p=0.817,=0.338),meeting the requirement of rapid and non-destructive detection of apple ripeness visual evaluation index.The transfer of the spectral model was basically achieved by using the third spline interpolation combined with the segmented direct correction algorithm,and the model correlation coefficient was 0.714 after the transferred model,and the root mean square error of prediction was 0.481.