Study On Cargo Segmentation And Indentification Technology For Intelligent Truck-loading Robots

This paper studies the method of cargo identification based on machine vision that can achieve high efficiency and high reliability,solves the problems of a series of job automation such as cargo loading,unloading,palletizing,and handing,and studies related software algorithm and software systems.The main study content of this paper is as follows:Firstly,according to the needs and the technical solutions?the work scenarios of the loading and loading robots,a structural scheme for using two SCARA robot arm for the loadingunloading module and the palletizing module is proposed.Secondly,in terms of cargo recognition,Kinect v2.0,which is released by Microsoft and can simultaneously acquire depth maps and color maps,is used as an image acquisition sensor.Moreover,Use the camera's pinhole imaging model and the calibration method that Zhang Zhengyou proposes to calibrate the Kinect's color camera and depth camera,and establish relative transform between color camera and depth camera.According to the working mode and environment characteristics of the loading-unloading manipulator and tray manipulator,the camera arrangements of Eye-in-Hand and Eye-to-Hand are adopted respectively.According to the structural features of three parallel rotations of the SCARA manipulators,a teaching-type hand-eye calibration method that use the transform among the manipulators,calibration object and the camera is proposed.Use the maximum likelihood method to optimize the calibration solution.The error analysis of the calibration result is performed through experiments and meet the requirements of the work accuracyThirdly,for the cargo with stable rectangular contour features and the rectangle has four L-shaped intersections formed by two mutually perpendicular adjacent edges,a method for cargo recognition by detecting rectangles is proposed.The usage scenarios and features of two algorithm based on contour detection and L-shaped intersection are analyzed from principle and experiment respectively.The detection algorithm that bases on the contours cannot accurately identify the goods whose boundary are not obvious,but the detection algorithm based on the L-shaped intersection point can just make up for this defect.Based on that,this paper proposes an integrated cargo identification method based on contour detection and L-shaped intersection detection to improve the robustness of cargo detection and verity its feasibility through experiments.Finally,we designed and manufactured an integrated model for loading and unloading mobile robots,and transplanted the software and algorithm into the robot prototype.The corresponding test experiments were conducted to verify the feasibility of the cargo identification method proposed in this paper and to meet the work requirements.