Study On Vis/Near Infrared Spectroscopy For Dynamic Online Nondestructive Detecting Sugar Content In Fruits

Now total area of fruit trees and fruit total output in China occupy the first place in theworld, but phenomenon of ‘difficult-selling’ often happen for its low process technique andthe higher demand quality. Sugar content is one of the major factors affecting taste in fruits.Routine analytical methods used for fruit sugar content measurement are destructive,expensive, complex, time and labor consume and off-line by nature. The technique ofVis/near-infrared spectroscopy analysis is the representative of “green measurementtechnique”, including efficient, rapid, low-cost, nondestructive characters. The main purposeof this study is to study Vis/near infrared spectroscopy for dynamic online nondestructivedetecting sugar content in fruits,, including the following four aspects:(1) A Vis/near infrared spectroscopy device was set up for dynamic onlinenondestructive detecting sugar content in fruits. The process involved optical systemconstruction, selecting spectroscopy, design of single conveyer for high-speed sortinginstitutions, setting up an automatic control system for dynamic position tracking anddevelops a software system for spectrum Analysis. The evaluation criterion was introduced tojudging online detecting precision for sugar content in fruits.(2) The influence of fruit size on online detecting precision for sugar content in fruitswas studied. Three sugar content models were established by different size. The sugar contentmodels of big, medium, small Gannan navel oranges gave the rvof0.81,0.79and0.86, theRMSEP of0.76,1.00and0.82oBrix, respectively. The sugar content models of multi-sizeGannan navel oranges gave the rvof0.84, the RMSEP of0.85oBrix. The findings suggesteddetecting precision for sugar content was significantly affected by fruit size. To improve therobustness of the models and to reduce the influence of fruit size on the models, a largenumber of different size samples are needed.(3) The influence of different stance of fruits on online detecting precision for sugarcontent in fruits was studied. The findings suggested that the stance of fruit had influence onthe results. The sugar content models developed by apples detected on the stance that fruitsequatorial plane paralleled to the running direction (CDPX), perpendicular to the runningdirection (CDCZ) and the stance that carpopodium faces to the detector (Top) gave the rvof 0.84,0.85and0.79, the RMSEP of0.58,0.62and0.90oBrix. The sugar content modelsdeveloped by Gannan navel oranges detected on the stance that CDPX and CDCZ gave the rvof0.73and0.92, the RMSEP of0.88and0.47oBrix, respectively. The multi-stance sugarcontent models developed by apples gave the rvof0.83; the RMSEP of0.69oBrix, by Gannannavel oranges gave the rvof0.75, the RMSEP of0.86oBrix, respectively. The results can beimproved slightly by the models. From the experimental results, fruits should be kept on thestance of CDPX or CDCZ for detecting the sugar content.(4) The influence of detecting speed on online detecting precision for sugar content infruits was studied. The findings suggested that the better models of sugar content in applesamples gave the rvof0.89, the RMSEP of0.69oBrix when the integration time was30msand the detecting speed was3per second. The highest precision could be made with thematching optimization parameters between detecting speed and integration time.(5) The multi-size and single-size models for detecting sugar content in fruits weredeveloped with the matching optimization parameters and pre-processing method SNV. Themulti-size calibration models gave the rvof0.92and the RMSEP of0.50oBrix. Thesingle-size calibration models gave the rvof0.92and the RMSEP of0.53oBrix. These modelsmeet the real-line detecting requirements.