Sorting Of Peeled Mandarin Segments With Orange Core Based On Fluorescence Characteristics

The existing machine vision system has a low accuracy in the sorting of the peeled mandarin segments with orange core.Therefore,this article analyzed the different optical properties between orange core and orange gizzard based on the difference of physical and chemical properties,to determine the maximization difference signal between orange core and orange gizzard as the conditions of image acquisition system.This will provide the accurate premise and foundation for the machine vision system to sort the peeled mandarin segments with orange core.Flavonoids have fluorescence properties,while the content of flavonoids in the orange core and gizzard is different,so the fluorescence properties of orange core and orange gizzard are different in theory.For the purpose of offering direct and accurate signals to machine to identify the peeled mandarin segments with orange cores.In this study,100 peeled mandarin segments were taken as the samples,the 3D fluorescence spectra of orange core and orange gizzard were determined by Cary Eclipse fluorescent spectrophotometer(at an excitation wavelength in the range of 300-700nm,5nm increments,an emission wavelength in the range of 350-750 m,1 nm increment,discharge voltage of 600 V,scanning speed of 1200 nm/min).Two-way ANOVA revealed the optimal excitation wavelength for detecting orange core was 370nm.One-way ANOVA was applied to the fluorescence emission spectra of all regions at 370nm excitation to determine the emission wavelengths for orange core detection.The major emission wavelength was 440nm.The differences of fluorescence intensities at 440nm for orange core and orange gizzard were evaluated by means of drawing the box-plots of the fluorescence intensity at 440nm.The accuracy rate of discrimination orange core and orange gizzard using the mean of lower quartile of orange core and the upper quartile of orange gizzard at 440nm as classification criterion.According to the classification criterion the discrimination accuracy of orange core and orange gizzard was 85.5%.The optimal excitation wavelength was selected as the light source of the hyperspectral image system.100 hyperspectral images of peeled mandarin segments with orange core were acquired in Gaia hyperspectral image system with the light source of(370±2nm).Emitted light from the sample was measured in 0.65nm intervals from 382nm to1019nm.ENVI5.1 was used to extract the spectral of the orange core and gizzard.One-way ANOVA analysis showed that the major emission wavelengths were 440nm?460nm and520nm.The machine vision system was configured based on the optimal excitation wavelength,the major emission wavelengths and the signal requirements of the AS-20 sorting system(light source 370±2nm,IMAGINGSOURCE DFK 33GP 1300 camera,filters 460±5nm.440±5nm?520±5nm).The images of peeled mandarin segments with orange core and peeled mandarin segments without orange core were acquired in the machine vision system,then put them into the AS-20 sorting system.The images were processed by grayscale,enhancement,binarization,morphology and so on,and the detection rate of the peeled mandarin segments with orange core was analyzed.The result showed that the detection rate of the peeled mandarin segments with orange core was 88.88%,with the binarization threshold of 75 and the filter of 460nm;the detection rate of the peeled mandarin segments with orange core was 88.75%,with the binarization threshold of 45 and the filter of 440nm.The monochromatic fluorescence image contains the exclusive signal of the orange core relative to the orange gizzard,which can be used for sorting the peeled mandarin segments with orange core by machine vision system.