| China’s apple planting area is the first in the world,but its export volume is very small,mainly because our country’s fruit post-harvest processing technology is relatively backward,it is difficult to meet the export requirements.At present,China basically relies on manual with mechanical way to classify,this sorting method,higher cost and low efficiency.The application of fruit quality online detection device for fruit sorting has the advantages of fast,non-destructive and high reliability.Near-infrared spectroscopy is widely used as a non-destructive and rapid detection technology for internal fruit quality sorting.Different parameters of NIR online inspection devices affect the NIR spectra,and the amount of spectral information required for fruits with different peel thicknesses is different,so the NIR spectral device parameters need to be optimized to meet the model accuracy requirements.As the NIR light propagates inside the apple,the energy carried by it will be attenuated,and as the apple fruit diameter(light range)increases,the more drastic the attenuation degree is,which will cause the variability of the NIR spectra and eventually cause the problem of poor performance of the apple soluble solids content prediction model.Therefore,there is a need to investigate the extinction rate of the internal pulp of apples and to establish an apple size correction and compensation soluble solids prediction model,which can be used to eliminate the influence of apple size differences on the soluble solids prediction model.In this study,the soluble solids content(SSC)of Fuji apples was used as the detection index to investigate the effects of different parameters of the NIR online detection device and the differences in apple fruit diameter on the prediction model of its soluble solids content,to improve the stability of the prediction model and to achieve optimization.The major findings and results of the thesis are listed below.1、Nine sets of spectra were collected under three experimental parameters using near-infrared spectroscopy online detection equipment.The performance of the apple SSC prediction model established by the experimental setup at different movement speeds,integration times,and light intensities was compared to optimize the parameters of the experimental setup.The results showed that the SSC prediction model built with a movement speed of 0.5 m/s outperformed that of 0.3 m/s at two different movement speeds.The SSC prediction model built with an integration time of 120 ms outperformed the other integration time groups at four different integration times.Finally,the performance of the SSC prediction model for apple built with four different light intensities was investigated under the experimental parameters of motion speed of 0.5m/s and integration time of 120ms.Various pre-processing and band screening methods were used for comparison,and it was found that the performance of the apple SSC prediction model established at a light intensity of 6.5A was generally better than that of the other light intensity groups.The reason is that when the light intensity is low,the performance of the apple SSC prediction model becomes better as the light intensity increases and the spectrometer detector receives more information about the interior of the apple.However,when the light intensity exceeds a certain value,as the light intensity of the halogen lamp increases,more stray light is brought into the spectrometer detector,and the performance of the established apple SSC prediction model becomes worse.The CARS band selection method was used to select the band for the experimental group with the light intensity of 6.5A,and the CARS-PLS model established by screening 83 variables from 1044 spectra was optimal,with R_pand RMSEP of 0.991 and 0.149,respectively.Parameter optimization of near-infrared spectroscopy online inspection equipment can help improve the performance of apple SSC models.It contributes to the provision of technical assistance for online sorting of apple quality.2、A fruit size-based transmission spectrum correction method is proposed to improve the performance of its sugar prediction model.The transmission spectra of apples of different sizes were converted to the same standard to obtain the extinction coefficients of transmitted light.The transmission spectra were corrected according to the average extinction coefficients of apples,and the corrected spectra were used to build the apple soluble solids content prediction model.Compared with the apple soluble solids content model without size correction,the modeling set in the 75 mm-85 mm fruit diameter group predicted the prediction set in the 65 mm-75 mm fruit diameter group,and its prediction set correlation coefficient R_pincreased from 0.769 to 0.869,and its prediction set root mean square error value RMSEP decreased from 0.990 to 0.721.The modeling set in the 75 mm-85 mm fruit diameter group predicted the prediction set in the 85 mm-95 mm fruit diameter group,and its prediction set correlation coefficient R_pincreased from 0.687 to 0.932,and its prediction set root mean square error value RMSEP decreased from 0.778 to 0.531.It is clear that the proposed apple size correction method is effective and can be used to eliminate the effect of apple size differences on the sugar prediction model.3、Apples were used to study the effect of seed diameter on the SSC pattern.The results show that apple size differences affect its spectrum,and the relationship between apple size and its spectral light intensity satisfies a logarithmic function,which eventually affects the prediction performance of the PLS model of apple SSC established by it.For this reason,the solution methods for different size differences were investigated,and a mixed model of apple fruit diameter and a mid-level fusion model of wavelength variables and apple size data after CARS screening were established.Comparing the different solution methods,we found that the size-compensated model for apple soluble solids content established by the intermediate fusion model of wavelength variables and apple size data after CARS screening had a prediction set R_pof 0.886 and RMSEP of 0.536 for fruit diameter of 65 mm-75 mm,and its prediction set R_pof 0.913 and RMSEP for fruit diameter of 85 mm-95 mm was 0.497,and the model performance was optimal.Therefore,adding fruit diameter variables and building a size-compensated model for apple soluble solids content can improve the accuracy of the proposed method.It satisfies the demands of detecting the SSC of apples with various apple fruit sizes online. |
PDF Full Text Request