| China has the highest yield and planting area of apple in the world. However, the export only accounts for below 3% of the production due to the low level of fruit’ commercialization and the detection technology. Therefore, considering the demand for exports and various needs of domestic consumers for apple product quality especially the internal quality,advanced nondestructive equipment must be developed to guarantee apple production management, and at the same time to sort the postharvest apples based on quality automatically, which will be quite helpful to improve the quality of apple and enhance market competitiveness.The soluble solids content(SSC) of Fuji apple of China is used as detection index. The research project is to use visible/near infrared(Vis/NIR) spectroscopy and hyperspectral imaging technology, to analyze the stability of spectroscopy and the calibration model and to develop a portable apple SSC Vis/NIR instrument. In addition, a regression method was proposed for the fusion of spectra and image information, and the detection of the surface distribution of apple SSC is also studied. Research contents and conclusions were listed as below:(1) The stability of spectral data was investigated. Diffuse reflectance and diffuse transmittance spectrum measurement system were designed for portable and on-line detection,respectively. As for diffuse reflectance measurement, spectra from different distances between light source and detector were analyzed. It was concluded that when the distance between light source and detector was 15-20 mm, the noise-signal ratio was good with the effective penetration depth of 20-30 mm, which was helpful for the development of the portable instrument. As for diffuse transmittance measurement, spectral area change rate was used to evaluate the stability of spectra collected in two fruit-orientations. Results showed that the stability of spectra from fruit-orientation(stem-calyx axis horizontal) was better. This method could also provide reference to evaluate other factors that influence spectral collection.(2) The calibration model for SSC was developed based on effective variables selectedfrom the combination of spectral data of apple samples of different origins, which has a great suitability for origins of apple samples. Apple samples from 4 different origins were used to establish local origin and hybrid origin models for the prediction of SSC. The results indicated that the model established based on 4 origins performed better than local origin and other hybrid origin models. With the help of 16 effective wavelengths, the prediction ability was improved with rp and RMSEP of 0.978 and 0.441°Brix, respectively. Then the model was evaluated by the apples from Changping, Beijing, with rp and RMSEP of 0.941 and0.509°Brix, which indicated that it could significantly reduce the origin effect on the robustness of Vis/NIR models for SSC of apple. It was found that the model developed by more samples of different origins combined with effective wavelengths showed good prediction ability for apple sample of unknown origin.(3) Selecting variables from the combination of spectral data of different measurement positions was proposed to reduce the influence of variation of spectrum measurement position on the robustness of NIR models of SSC. The spectra were collected around stem, equator and calyx positions and then used to develop global models. With the help of 66 effective variables, the global model yield more promising results with RMSEP for three measurement positions being 0.409, 0.386, 0.486°Brix, respectively. Moreover, another spectral data from equatorial position were used to develop SSC calibration models with full-spectrum and effective variables, respectively. Compared to the model with full-spectrum, the model based on 66 variables was well suited to control the prediction accuracy for SSC with the RMSEPs for 3 different positions decreased from 0.779, 0.661, 0.682°Brix to 0.645, 0.530, 0.614°Brix.The overall results indicated that the global model built with effective variables could make the variation of spectrum measurement position a negligible interference for SSC prediction and the effective variables selected could provide a reference for offsetting the effect of spectrum measurement position variation on models established previously only using the spectra acquired from the equatorial position.(4) Hyperspectral imaging was used to detect the SSC of apples. A new regression method called combined partial least square(CPLS) was proposed to analyze the integrations of spectra and different kinds of textural features which all extracted from apple hyperspectral images. The prediction results indicated that the CPLS model developed with the integration of spectra and the contrast feature at angle of 0°(contrastθ=0) achieved promising results(RMSEP=0.596°Brix) and significantly improved SSC prediction compared with the spectral data when used alone(RMSEP=0.621°Brix). Then the spectral data from stem, equator and calyx positions were extracted and used to develop calibration model after pretreatment by 1st derivative and SNV. The model could predict the SSC with better results for 3 differentpositions with the RMSEPs of 0.666, 0.623 and 0.645°Brix. Last, the model was used to predict every point of apple hyperspectral images. The results showed that the prediction results meet the actual distribution of SSC, but the prediction error was a little higher when the model was used to predict the portion of edge of apple. The detective precision still needs to be improved to achieve the final purpose.(5) According to above-mentioned research, a portable apple SSC Vis/NIR instrument was designed with a ring fiber probe, a portable spectrometer and a WinCE embedded operating system. Calibration model of SSC prediction was developed and modified with effective variables. Then the instrument and its prediction model were validated by new validation samples with rp and RMSEP of 0.910 and 0.579°Brix, respectively. These results indicated that the instrument could meet the needs of SSC detection conveniently and rapidly,but the stability and reliability of the system still needs to be further improved. |
PDF Full Text Request