Study Of Outdoor Illumination Influence On Wonhuwang Pear VIS/NIR Acquisition

Visible/Near-infrared(VIS/NIR) spectra analysis technology, as a rapid and non-destructive technique, now is widely used in fruit quality determination.However, this technology mainly research and application environment is only in the indoor constant, while is not widely used in the outdoor. In this paper take the Wonhuwang pears as the subject, and research the outdoor illumination effect on near infrared spectrum detection. To establish the mathematical model between the sugar and spectrum, using a transfer model reduce the gap between the indoor and outdoor spectrum. Also, this paper was used a non-linear model to reduce the influence of outdoor illumination. The results and conclusions were listed as follows:(1) Using the Zeiss visible/near infrared spectrometer to obtain the spectra of Wonhuwang pear under two kinds of lighting conditions: indoor and outdoor. Then,build a PLS model between spectra and the sugar, and using the PLS model to mutual predict spectra under different conditions. And then discuss effect of the band selection and pre-treatment on modeling. The experimental results have three aspects.First, the indoor spectrum model directly to prediction is not better, and it is not suitable for application in fact. Secondly, the band selection in near infrared region is better to modeling, this model of the correlation coefficient R, RMSEC and RMSEP were 0.78, 0.36 and 0.61. The last, the first derivative(FD) pre-treatment is better,this model of the correlation coefficient R, RMSEC and RMSEP were 0.75, 0.489 and 0.536.(2) Using the Wonhuwang pear indoor spectrum as master instrument and the outdoor spectrum as slave instrument. Selecting the standardization sample number was by KS algorithm and SPXY algorithm. Also, applying S/B algorithm, DS algorithm and PDS algorithm carried out the transfer model of the outdoor spectrum.The result indicated that the transfer model could effectively solve the gap between the master instrument and slave instrument. In other words, it can effectively reduce the influence of illumination. The model is better by DS algorithm. And this model of the correlation coefficient R and RMSEP were 0.7 and 0.374.This model couldeffectively close the spectral gap between indoor and outdoor by DS algorithm.(3) Using four kinds of uncommon algorithm to build the model: step multiple linear regression(SMLR), partial least squares(PLS), genetic algorithm partial least square(GA-PLS) and least squares support vector machine(LS-SVM).The results showed that the uncommon algorithms are also to reduce the measurement gap under different illumination. However, the SMLR algorithm and the GA-PLS algorithm are not suitable for use in complex environment.