Using NIR Spectroscopy In Nondestructive Determination Of Navel Orange Quality Based On Android

Navel orange is a kind of popular fruit which is accepted by customers for its essentialnourishment such as a good source of vitamin C and carotene. China has become the biggestacreage and producer country of navel orange in the world, but there exist great shortage instandardization process and quality control. The development of portable device is not onlyapplied to detect quality of postharvest fruit, but also assessed quality at the growing season.So it is necessary to develop a portable device to accomplish fruit quality determination.In this paper, the design of portable device based on android is used to measure the sugarcontent, pH and epidermal thickness of common Newhall and Washington navel orangeswhich were produced in Jiangxi province by near infrared diffuse reflectance spectra. Inspectral range of950nm~1650nm, this research systematically studied the NIR spectralcharacter and the influences from various modeling analysis techniques such as pretreatment,variable selection and modeling method on the accuracy of predictive models. Through acomprehensive comparison, the optimal method and the best model was obtained.The main contents and conclusions of this research were as follows:(1) The characters of MicroNIR1700spectrometer was introduced comprehensively.Further, the portable stable device which determined quality of fruit was designed. The devicewas presented in detail and processed.(2) The software of spectral collection and prediction of fruit’s quality was programmedwith Eclipse and Android SDK. The main function of the software which widely used asuser-friendly control and multiple functions are spectral collection, parameter setting, qualitypredicting and so on.(3)The quality models of navel orange were built. The partial least squares (PLS) modelsfor SC and pH of navel orange with two kinds of variety and different periods obtained withdifferent pretreatment methods and different variable selection methods such as geneticalgorithm (GA), successive projections algorithm (SPA) and so on. For SC of immatureNewhall and Washington navel oranges, the prediction correlation coefficients (Rp) were0.896and0.839, the root mean square errors of prediction (RMSEP) were0.212and0.208; For pH of immature Newhall and Washington navel oranges, Rpwere0.724and0.796,RMSEP were0.105and0.073; For SC of mature Newhall and Washington navel oranges,Rpwere0.825and0.933, RMSEP were1.138and0.994; For pH of mature Newhall andWashington navel oranges, Rpwere0.745and0.772, RMSEP were0.112and0.111.(4) Chemometrics methods combined with near infrared diffuse reflectance technology were used to establish PLS model in detecting epidermal thickness. The model optimizationswere carried out by GA and these results were better than other variable selection methods.The best calibration models built with PLS were obtained with Rpof0.864and0.875, andRMSEP of0.290and0.257for two kinds of navel oranges.