Research On The Robot Sorting System Of Construction Waste Based On Vision

With the rapid development of infrastructure construction and urbanization in China,the scale of urban construction is increasing year by year.However,urban construction produces a large amount of construction waste,which causes serious pollution to the ecological environment.Therefore,the resource utilization of construction waste is imperative.At present,the construction waste disposal equipment has realized the automation of part of the process,in which the separation of wood and plastic still adopts the manual sorting method.In order to realize the automatic sorting of this part of the process,the traditional visual algorithm is difficult to identify,classify and locate the construction waste due to its various types and complex shapes,so the YOLO target detection algorithm sorting robot system based on the convolutional neural network is proposed.For sorting task requirements,design the sorting system overall scheme,for sorting robot kinematics analysis and simulation,according to the Zhang's method inside and outside the camera parameters and hand-eye calibration experiment,completed the calibration of the visual system,established the target object in the robot base coordinates posture relationship,YOLO algorithm based on the target object recognition and localization are analyzed,by sorting experiments,implementation of the construction waste sorting of wood and plastic.Research contents include:1.The D-H method is used to model the industrial robot,and the forward and inverse kinematics are solved,and the correctness of the forward and inverse solutions of the robot is verified by simulation.This paper analyzes the trajectory planning method of cubic and quintic polynomial interpolation,establishes the industrial robot model in Matlab robot toolbox,carries out the simulation analysis of trajectory planning on the robot end,and obtains the smooth and continuous angular displacement,velocity and acceleration curves of each joint axis.Then,in the vision system,Zhang's method is used to calibrate the camera,to obtain the internal and external parameters of the camera,and then through the hand-eye calibration experiment,the transformation matrix of the camera in the robot's base coordinate system is solved.2.In terms of target recognition and positioning,the network structure and detection principle of YOLO target detection algorithm based on convolutional neural network are analyzed.Construction waste images are collected as data sets,and the image data sets are expanded with data enhancement technology,and the target objects in the data sets are marked,so as to train my YOLO detection model.In order to test the image conveniently,a construction waste target detection system based on YOLO algorithm is designed.Using the detection system to identify construction waste,YOLO model can accurately detect the position,category and confidence of the target object in the image.3.Design an experimental platform composed of industrial cameras,computers and industrial robots to verify the effectiveness of YOLO detection algorithm.Design the robot sorting experiment,through the work on the stage of the construction waste sorting for 50 times experiments,among them,the recognition rate of lumber and sorting the success rate of 96% and 90% respectively,plastic recognition rate and sorting the success rate of 94% and 92% respectively,the experimental results show that the sorting system based on YOLO target detection algorithm can effectively identify and locate the target object,and through the robot to realize the target object sorting,verify the feasibility of the system.