Reseaerch On Key Technology Of Visual Inspection And Recongnition In Elevator Hall Door Robot Packing System

Automation,informatization and intellectualization have become an inevitable trend in the development of manufacturing industry.As one of the ten strategies of "made in China2025",industrial robot technology represents the future development direction of intelligent equipment.Industrial robots are made more intelligent by adding visual sensors and force sensors to perceive certain states in the process of operation and forming feedback,as well as adding algorithms such as artificial intelligence and deep learning.The identification and classification of workpiece is the main application form of industrial robot in production line.At present,most of the robots widely used in industrial production lines complete preset fixed actions and functions through instructional programming or offline programming.Once the workpiece does not match the preset position,the robot will not be able to complete the task and even may be damaged.In this paper,the robot packing sheet element in a flexible production line of elevator hall door was used as the experimental platform to study the problems of packing failure in the production process and dumping of the hall door after packing.By adding visual system to industrial robots,the intelligence of robots is improved to solve the problems in the process of production.The main research contents of this paper are as follows:This paper introduces the function,composition and production process of the flexible production line of elevator hall door,analyzes the reasons for the problems of the robot packing unit in the production process,and puts forward the visual detection and recognition technology solution based on the robot hand-eye system.This paper studies the visual positioning of elevator door packing cases and the flexible packing technology of robot.Firstly,the camera’s internal and external parameters are obtained by using zhang zhengyou’s visual calibration algorithm.Then,the kinematics analysis of the end-effector of the packing robot is carried out,and the hand-eye calibration of the packing robot is completed.Finally,the positioning packing box image information collected by the visual system is processed to obtain its positional angle,and the robot adjusts the end-effector posture of the robot in real time to complete the hall door packing action.In this paper,the classification technology of elevator door packing state is studied:firstly,the state recognition of elevator door based on support vector machine is studied,and the classification algorithm of support vector machine is designed.Then,the convolutional layer,pooling layer and activation function used in the convolutional neural network structure are studied and analyzed,and the corresponding deep learning convolutional neural network structure is designed.Through comparative experiments,it is verified that the deep learning based state recognition algorithm has higher accuracy and better robustness under the interference of many factors in the actual industrial working environment.Finally,experimental verification is carried out.The experimental results show that the vision-based flexible packing technology of elevator door robot and the identification and classification technology of elevator door packing state are feasible,stable and can meet the normal production requirements.