Research And Application Of Desktop Robot Arm Based On Machine Vision

With years of research and development of the robotic arm in the industrial field,many breakthrough results have been obtained.Industrial manipulators are widely used in work and production due to their high efficiency and low cost.Unlike the more closed and monotonous working environment in the industry,the intelligent interaction of the desktop robotic arm in the human-machine environment is facing many challenges.Desktop robotic arms are mostly used in homes and teaching environments.The environment has high complexity,diverse object shapes,and different sizes,with strong uncertainty.In complex scenes,multiple target objects are arbitrarily stacked and placed,which places higher requirements on the intelligent recognition and grasping of the robotic arm.Compared with ordinary industrial robot arms,desktop robot arms have the characteristics of small space occupation,low price,and human-computer interaction.In order to better realize the man-machine interaction of the robot arm,the main research contents and results of this article are as follows:1)This subject mainly adopts modular design,with Raspberry Pi as the core control board,designing middleware code and motor control board.Ensure that the robotic arm moves accurately,and grasp the reagents on the desktop of the chemical laboratory.2)Add a chemical reagent bottle identification and positioning function to the desktop robotic arm: At present,there are many application frameworks in production and life to achieve machine vision.According to the characteristics of the small working space of the chemical laboratory table and the complexity of the reagent placement space,a suitable visual platform is built.It is the focus of this study to improve the system recognition rate and ensure that the system accurately and efficiently completes the actions on the reagent bottles.This subject mainly uses the Faster R-CNN regional target detection algorithm to realize the rapid recognition of chemical reagent bottles by the system,combined with Zhang's calibration method and hand-eye calibration method to achieve the system's distortion correction of two-dimensional image parameters and solve the image coordinate system to the machine The transformation of the arm coordinate system controls the mechanical arm to accurately grasp the target.Tests show that the target recognition method based on deep learning neural network is superior to the target recognition method using OpenCV in terms of recognition efficiency.3)Add reagent information classification query function to the desktop robotic arm: realize the identification,classification,storage and provide related network information query function for the reagent bottle information in the working environment of the robotic arm.By writing a Python intermediate program and combining Baidu AI text recognition,the identification of the chemical label text information in the target area of the picture is completed;the network information of related reagents is obtained through the intermediate program and stored in the cloud MySQL database,which effectively improves the storage efficiency of information;Provide target reagent information retrieval function to facilitate the user's unified query and management of related reagent information,improve the efficiency of researchers,and ensure the safety of the experimental environment.Tests show that the system can quickly identify target information and the accuracy can reach about 96%,which is convenient for users to realize the information management of laboratory chemical reagents.