| The key point that reduces the loss rate of fruits and improves their quality is by determining the harvest date. Earlier or later harvest will affect commercial value of fruits. The quality of fruits depends on the fruits varieties, climate conditions, and origins, so it's really difficult to determine when to harvest. At present, orchardists generally chose the harvest date based on their experience. With the large-scale and automatic development of agricultural facilities in China, it poses an urgent need in developing instruments and methods for field test, so as to provide a scientific basis for determining the harvest date.In this study, Huanghua pears were chosen as the object, and the detection method and equipment for determination of harvest day were developed using Visible/Near Infrared Spectroscopy (Vis/NIRS). Characterization factors of the optimal harvest day of Huanghua pears were selected by factor analysis, and an instrument for spectral measurements in field at harvest time was built, meanwhile, discriminant models of harvest day and detection models of SSC content were established, the influence of tree shapes, years and harvest time on these models was discussed. Main contents and conclusions of this research were listed as below:(1) Spectral characteristics of Huanghua pears in different harvest days were studied. With self-made integrating sphere as diffuse reflection accessory, an instrument for spectral measurements was set up. Through analysis of six different harvest time spectra of Huanghua pears, the results show that the main absorption peaks of Huanghua pears' spectra are around 680,970,1190 and 1450 nm, Among which,680 nm belongs to chlorophyll's absorption bands. Before and after the optimal harvest time, Huanghua pears'skin changes from greenish-brown to yellowish-brown, the content of chlorophyll decreases while the reflectance increases. Meanwhile, with the extension of harvest time, reflectance shows a downward trend at the remaining three absorption peaks, which is mainly caused by the absorption of carbohydrates in fruit.970 nm and 1190 nm belong to the second overtone and the third overtone regions of C-H, which are closely related to the absorption of sugar in fruit. In addition, the absorption peaks of fruit hardness are mainly around 1100 nm and 1390 nm. The results show that the change of main compositions in Huanghua pears during harvest time can be fairly reflected by VIS/NIR spectra, which provides a basis for the usage of spectral analysis technology in the harvest time determination.(2) Factor analysis was adopted to select the characterization factors of Huanghua pears during harvest time. In this study, nine items of basic information were collected and analyzed from 432 Huanghua pears in the opening canopy trees at 6 different periods of harvest time, such as the diameters, the hardness, SSC, pH, SSC/pH, etc. We find that the cumulative contribution rate of the first three factors is 79.5%. A comprehensive evaluation index F for Huanghua pears' maturity evaluation was established. And then the SSC content, all original basic variables and the comprehensive evaluation index F were taken as characterization factors of fruits maturity, respectively. The correct recognition rates of 432 Huanghua pears at 6 different periods of harvest day is 94.29%,68.75%,55.56%,64.94%,46.75% and 65.08%; 97.14%,76.56%,60.98%,64.94%,73.42% and 77.78%; 70.00%,42.19%,64.20%,42.19%, 77.92% and 80.95%, respectively, which indicate that the result of correct determining the harvest day of Huanghua pears based on the comprehensive evaluation index (F) is more effective. Then the spectral calibration model of the index F was built by PLS, with determination coefficient of 0.91 and RPD of 3.38. And the good prediction of the model provides a basis for the usage of the spectral analysis technology in selecting optimal harvest day for Huanghua pears.(3) Discriminant models of harvest day for Huanghua pears were built based on the comprehensive evaluation index (F). At the beginning, the models were set by using traditional PLSDA method. The results show that the recognition rate of the model for single tree shape fruits in optimal harvest time predication of fruits of the same year and the same tree shape is 100%, while the recognition rate for fruits of the same year and different tree shapes is between 30% and 57.9%, and the recognition rate for fruits of different years is 0, which indicates that discriminant model has a poor applicability for fruits of different years and trees. Then the spectral calibration model of the index F was further used, combined with the interval distribution of F at different harvest time to determine the optimal harvest day for Huanghua pears. The results show that the recognition rate for fruits of the same year and different tree shapes is 78.95% and 73.68%, respectively. And the recognition rate for fruits of different years and different tree shapes is 66.76% and 50%, respectively. Above all, the discriminant model based on the comprehensive evaluation index (F) is very practically valuable, which eliminates the effects of tree shapes and years.(4) The spectral detection model for SSC content of Huanghua pears at optimal harvest time was studied. Meanwhile, the influence of tree shapes, years and harvest times on robust and accuracy of the detection model was discussed and primarily evaluated. The accumulation of sugar in fruits were affected by the light permeability of trees, so the quality of the fruits varies as the shape of the trees, thus the applicability of the model was affected. The detection model of SSC content of Huanghua pears predicted more precisely in single tree shape of a single year, with the root mean square error of prediction (RMSEP) from 0.40 to 0.58°Brix and RPD of more than 2.5, which suggested it was suitable for quantitative analysis. But its root mean square error of prediction (RMSEP) ranged from 0.63 to 0.74°Brix and RPD from 0.90 to 2.22 when predicting spectra of fruits of single year in different tree shapes, and the prediction precision and robustness decrease. Meanwhile, the root mean square error of prediction (RMSEP) of combined model of fruits of single year in different tree shapes ranged from 0.45 to 0.54°Brix and RPD from 1.54 to 3.09. Thus detection model of SSC content of Huanghua pears in single year was built. When such detection model was used to predicate fruits in different years, its root mean square error of prediction (RMSEP) was from 0.95 to 1.23°Brix and RPD was less than 1.5. So the model can't be used in SSC content test for different years. And the root mean square error of prediction (RMSEP) was 0.56°Brix and RPD was 2.41 when detected by joint detection model of SSC content based on two different years. The detection model was affected by the frequencies of harvest time in the calibration set. With the increase of the frequency of harvest time in the calibration set, the calibration model became more applicable, predictable and robust. The RPD of the model was 0.95,0.75, 2.56 and 0.85,1.03,1.54, respectively when the detection model was built based on one period, three periods and five periods of harvest time in two different years. Therefore, the tree shapes, years and harvest time are decisive to the precision and robust of the detection model of SSC content of Huanghua pears in optimal harvest time, which will be valuable for further study in the application of spectroscopy in fruits in field.(5) An instrument for spectral measurements of Huanghua pears during optimal harvest time was developed in field. Normally, it was difficult to collect the effective spectral of pears in field due to the strong interference of ambient light. In this paper, a method of ambient light correction was proposed. In order to reduce the effect of ambient light, a shutter was attached to the front of the probe. When the shutter was opened, the spot spectra will be obtained under the influence of the instrument light and ambient light. Then the background spectra which were only influenced by the spot light will be obtained when the shutter was closed. Then the background spectra will be subtracted from the spot spectra to do the ambient light correction. So the shutter was introduced into the instrument for spectral measurements. And the result shows the background correction could efficiently reduce the effect of ambient light on spectra data. But when the samples were exposed directly in the sunlight, there would be a strong interference between the background light and the instrument light inside the fruit. So it is just necessary to cut out the direct sunlight when the background correction method was used.(6) A comparative analysis of detection model of SSC content of Huanghua pears under different light conditions was carried out. Huanghua pears' spectra were collected both in field (in the shade) and indoor. And the correction model of SSC content was built based on the spot spectra in field, spot spectra after background correction, and indoor spectra, respectively. The root mean square error (RMSEP) of model is 0.87°Brix,0.59°Brix and 0.35°Brix, respectively, and the RPD is 0.79,1.69, and 2.58 correspondingly. Meanwhile, the root mean square error (RMSEP) of the prediction is 0.89°Brix,0.42°Brix, and 0.27°Brix, while the RPD is 0.79,1.69,2.58 correspondingly. Then the spot spectra and spot spectra after background correction were introduced into the indoor model, which produced the root mean square error of prediction (RMSEP) of 4.32°Brix and 0.55°Brix, respectively. And the results indicated that the background correction method and the equipment developed hereby were useful, but its accuracy need to be further improved. |
PDF Full Text Request