Computer Vision-based Intelligent Waste Bottle Classification System

With the rapid growth of the world population and the rapid development of the global economy,the limited resources have become an important factor limiting human survival and development,and the classification of waste is an important prerequisite for resource recovery.Waste bottles are an important recyclable resource with the following characteristics:(1)A large number: According to the British Health Bureau,the global plastic bottle consumption in 2016 reached 480 billion,and it is expected to reach 500 billion in 2020,equivalent to 20,000 plastic bottles per second.(2)High recycling value: Waste bottles can be divided into plastic,iron,aluminum and glass.Waste plastics,scrap metal and waste glass belong to China's top ten renewable resources.(3)Different types of waste bottles have different follow-up treatment methods: plastic bottles are treated by compression molding,extrusion molding,injection molding,etc.,while scrap metal bottles such as iron and aluminum are processed by compression,grinding and polishing.Therefore,the waste bottle classification system has high research value.At present,there are few related researches on waste bottle sorting systems,and their applications are mainly divided into traditional trash cans and smart trash cans.There are still some shortcomings:(1)In the early stage of the traditional garbage can,it is simple to arrange two or more barrels and identify recyclable and non-recyclable.It mainly relies on post-collection and classification,so that although it can achieve higher accuracy,it has low classification efficiency.Long classification period,cumbersome manual sampling and high labor intensity.(2)The intelligent trash can binds the garbage to the relevant responsible persons through automatic weighing,voice recognition,mobile scanning and other technologies,and improves the garbage sorting efficiency by supervising the user in the previous stage,but there are many types of sensors.It is expensive and has high technical requirements for operation,operation and maintenance personnel.This paper mainly proposes a computer waste-based intelligent waste bottle classification system,which studies the shortcomings of the current waste bottle classification system:(1)Design lightweight hardware devices: use as few sensor modules as cameras,servos,and light strips to provide hardware support for the system,reducing hardware costs and maintenance costs.Use the Raspberry Pi,Arduino development board and Android Things operating system to realize the overall process control of the system to ensure efficient,smooth and stable system operation.(2)Designing a variety of software systems: divided into mobile APP and Web-side background management system.Based on the Android system,the mobile APP is divided into a mobile user APP and a mobile merchant APP,which provide users with rich credit redemption functions and provide merchandise sales platforms for merchants,improve user participation,and promote merchandise trade.The web-side background management system is developed using the SSM(Spring+SpringMVC+MyBatis)framework to provide administrators with a fully functional data management platform.(3)Accurate and efficient recognition algorithm: It is divided into two recognition modes: barcode recognition and neural network recognition,which improves recognition efficiency and accuracy,and improves the diversity of recognition modes.Barcode recognition mainly uses the OpenCV library to detect whether a waste bottle image contains a barcode and uses the ZBar library to identify the content in the barcode.The neural network recognition uses the TensorFlow framework to identify the network based on the InceptionV3 model using migration learning technology,achieving a recognition rate of about 97%.