Research And Implementation Of Industrial Robot Sorting System For Municipal Solid Waste

With the acceleration of urban development in China,the amount of urban garbage storage is also increasing,ensuring high-efficiency garbage recycling is a top priority.In the traditional waste sorting line,there is usually a manual sorting platform.Relying on manpower to complete the task of garbage sorting,not only the operation efficiency is low,but also there are security risks,so this traditional garbage sorting method needs to be improved urgently.Industrial robot not only saves workers from repetitive labor,but also reduces production cost and improves sorting efficiency.But the shape of solid waste in the sorting line is complex and disorderly,so it is difficult for the sorting system to accurately identify and locate it,and it is difficult for industrial robots to be widely used in the sorting task.In order to solve the above problems,this paper introduces deep learning technology to realize the accurate recognition and positioning of solid waste objects,ensuring that industrial robots can accurately grasp the target when performing the sorting task.The research in this paper involves three key points: camera calibration,solid waste identification and positioning under deep learning technology,and building an industrial robot sorting platform for experimentation.The main work results of this article are as follows:(1)This paper studies the principle of camera imaging,compares the linear model and nonlinear model of camera,designs a hand eye calibration method suitable for industrial robot to grasp solid waste objects,and realizes the purpose of mapping the coordinates of image pixel points and robot coordinate system.(2)A convolutional neural network model—PDN model and KPPN model—suitable for solid waste sorting tasks at industrial sites is designed.The solid waste detection and key point location are realized in two steps.First,the solid waste is detected by the PDN model,and the position information of the input box is input.Second,the KPPN model uses the supplementary frame information as input to output the key points of the solid waste.The performance of the model is tested under different experimental conditions.The test results show that the two models can accurately detect and locate the solid waste objects.(3)Establish an industrial robot sorting platform in the real environment of solid waste sorting,and conduct robot sorting experiments under different experimental conditions.Finally,experiments show that the model proposed in this paper has certain robustness and effectiveness.The robot sorting system can be applied to the industrial sorting site with complex environment.