Research On Vision Driven PET Oil Bottle Identification And Intelligent Control System

PET bottle recycling is a key part of solid waste treatment.Recycling of waste plastics is becoming increasingly important as the problem of plastic pollution grows each year.PET bottles can become valuable recycled materials after being decalcified,cleaned and crushed.When oil bottles are mixed in the recycling process,it is necessary to identify the oil bottles first and distinguish them from non-oil bottles before they can be crushed,otherwise oil leakage will occur,causing pollution of recycled materials and affecting product quality.Due to the single repetitive process of manual sorting and recycling of PET oil bottles,the labour intensity is high and the identification accuracy is low,making the recycling of PET oil bottles inefficient and slow.To solve the above problems,an intelligent oil bottle identification and sorting control system combining machine vision and robotics is designed to improve the accuracy of the PET oil bottle recycling system,focusing on overlapping PET bottles,colour classification,oil bottle identification and bottle grasping,etc.The details of the research are as follows:Kinematics and dynamics models for oil bottle grasping robot are developed.Robot kinematics and dynamics studies and kinematic simulations are completed.The calibration of the vision system is achieved,and the image ROI acquisition is completed by studying the least squares fitting adaptive threshold segmentation method.Overlap recognition,colour classification and oil bottle recognition image processing algorithms are proposed for PET oil bottle recognition.The Hu moment contour matching method based on Gaussian pyramid,the SVM colour classification algorithm based on PSO and the Do G algorithm based on contour refinement are investigated according to the characteristics of PET bottles and the requirements of the recognition industry.The contour matching of overlapping PET bottles,the construction of PET bottle color classifier and the optimization of oil contour edge detection algorithm are completed.The accurate identification of overlapping bottles,the classification of bottle color and the accurate identification of oil bottles are realized.A control scheme for oil bottle grasping robot is investigated.According to the requirements of PET bottle intelligent grasping control,a fuzzy adaptive inversion controller control scheme is determined to reduce the influence of robot model uncertainty and external interference on the control system.The fuzzy adaptive inverse controller model is developed with Simulink.Experiments on trajectory tracking are carried out on a simplified two-joint robot model.The validity of the proposed algorithm is verified.The software for intelligent identification and gripping control of oil bottles is developed and the experimental platform of PET oil bottle identification and intelligent control system is built.Qt is used to write the host computer software,and the communication between the host computer and the PLC is completed according to the TCP/IP protocol,the communication between the ABB robot and the PLC is completed through the PROFINET protocol.In respect of the required functions,image processing programs,PLC control programs and robot control programs are written.The viability of the method in this paper is verified via the experimental platform.