Abstract: In this study， a plastic infrared sorting machine was used to conduct sorting tests on the normal polyethylene terephthalate （PET） plastic bottles with blue， white， and green colors as well as the other plastic bottles made from polycarbonate （PC）， polyethylene （PE）， and polypropylene under the dual-path excitation environment of visible and infrared light. Three common Retinex color enhancement algorithms were further utilize to identify the white-， blue-， and green-color PET bottles， and the recognition effect of plastic infrared sorting machines was investigated. The results indicated that there was no significant difference in the imaged images between PP， PC， and PET bottles with a white color under visible light. However， under infrared imaging， PP， PC， and PET bottles reflected green， purple， and blue light with significant differences. The HSV range value was expanded from 0~46 to 0~61， the background saturation was adjusted from 72 to 80， the missing proportion of white and blue PET bottles was reduced from 4.58 % to 1.26 %， and the misidentification rate of white bottles made from PP， PC， and PE was reduced from 3.88 % to 2.69 %. There was no impact on the misidentification rate of the PE bottle， maintaining a recognition rate of 100 %. The best removal effect was obtained through the MSR color enhancement algorithm with removal rates of 94.8 %， 95.83 %， and 95.83 % for the recognition of white， blue， and green PET bottles， respectively.

Key words: infrared light, visible light, visual inspection, plastic identification, Retinex algorithm