Research On Human Posture Estimation Method Of Safe Operation System Of Industrial Robot

With the transformation and upgrading of our country’s intelligent manufacturing industry,the demand for industrial robots is increasing.Human-computer interaction refers to the process of cooperative operation between operators and robots,which is a major trend development of industrial robots.Because the safety of industrial robots is difficult to effectively ensure when cooperating with people,there is an unpredictable situation in the process of human-computer collaborative work,and what is more serious is the injury to operators.In order to ensure the safety of the staff in the human-robot collaboration system,this paper studies the human posture estimation method for the industrial robot safe operating system.Taking the human body of the operator during the operation of the industrial robot as the research object,the depth camera is used to collect the human body image information,and the key points of the human body posture are detected to future identify the human body posture,provide reliable information for the industrial robot to avoid obstacles,and improve the industrial robot operating system security.The main contents are as follows:Firstly,the depth camera is used to collect the human body image information of the operator.For the key points of the human body in the image information,a lightweight openpose network detection method is proposed to detect.The network proposed in this paper can reduce the complexity of the model while maintaining the accuracy.The experimental results show that the network can detect human key points faster.It lays the foundation for subsequent human pose estimation and recognition.Secondly,in view of the problem of some misidentified points in the acquired human pose key point data,a human pose estimation method based on Kalman filter is proposed,and a human pose long-term memory network(LSTM)model recognition algorithm is established.The Kalman filter is used to process the coordinate data of key points of human posture to eliminate the influence of abnormal points,and then define the target posture data set to establish a human posture LSTM network model to classify and identify the human posture.The proposed LSTM algorithm has a good recognition effect.It can classify and recognize more poses to establish the LSTM network model of human poses.Finally,the human body image information collection,human pose estimation and recognition are encapsulated,and a human pose estimation and recognition system based on the ROS framework is built.Moveit control module and robotic arm obstacle avoidance module.Using the distributed communication framework of the ROS system,each module communicates through the ROS Topic and the Node to realize the design of the ROS-based human pose estimation and recognition system to meet the needs of human-computer interaction.