Identification for Recycling Polyethylene Terephthalate (PET) Plastic Bottles by Polarization Vision

Zhiying Tan,Zilong Wang,Zhongwen Fei,Xiaobin Xu,Jian Yang,Baolai Zhao

doi:10.1109/access.2021.3050816

Abstract

The recycling of disposed PET plastic bottles has significant meaning for environmental pollution control and resource recycling. The reuse value heavily depends on the purity of the recycled PET. In order to improve the purity of PET recycling, a new means of material identification of PET plastic bottle based on polarization vision is proposed. To identify PET plastic bottles on conveyor belt, an effective region segmentation method for plastic bottle location and feature extraction has been put forward. The polarization of the scattered light on the surface of the object contains the information of its material. To obtain robust material characteristics of the PET plastic bottle, an approximate refractive index calculation method based on the degree of polarization and phase angle is introduced. The mean, standard deviation and coefficient of variation of the approximate refractive index and intensity distribution are used to construct the feature vector. The support vector machine (SVM) algorithm was used for PET identification. The experimental results show that the mean accuracy of PET plastic bottles material identification of recycled bottles is up to 92.06%.

Highlights

Due to the strong barrier property, PET has become one of the most important packaging material for almost all kinds of beverages, including juice, milk and its products, tea, mineral water and some foods
For identification and separation among containers and bottles made of PET, HDPE, polyvinyl chloride (PVC), PP and PS, a ‘‘Two-Filter’’ identification method was proposed based on Near infrared (NIR) reflectance spectroscopy [17]
Influenced by the ambient light and the noise introduced by the sensor, there is some noise in the calculated approximate refractive index of the plastic bottle

Summary

INTRODUCTION

Due to the strong barrier property, PET has become one of the most important packaging material for almost all kinds of beverages, including juice, milk and its products, tea, mineral water and some foods. The purity of the recycled PET depends on the identification rate of PET bottles out of many different plastic products. The magnitude and polarity of the generated charge depends on the contacted polymers material can be used sorting plastics into PP, PET/PS, PVC and HDPE, it is difficult to distinguish PET from PS [9]. Z. Tan et al.: Identification for Recycling PET Plastic Bottles by Polarization Vision. For identification and separation among containers and bottles made of PET, HDPE, PVC, PP and PS, a ‘‘Two-Filter’’ identification method was proposed based on NIR reflectance spectroscopy [17]. In order to obtain a higher identification accuracy of recycling PET plastic bottles, a method combining polarization vision and support vector machine (SVM) algorithm is proposed. A method to calculate the approximate refractive index of recycled plastic bottle based on the degree of polarization and azimuth is introduced. SVM is used to train and learn the feature vector, and the recognition model is obtained, which is verified by experiments

SCHEMATIC OF PET IDENTIFICATION FOR PLASTIC BOTTLES

EXTRACTION OF FEATURES

EXPERIMENTAL RESULTS AND DISCUSSION

CONCLUSION