Research On Kinect-Based NAO Robot Motion Imitation

With the rapid development of artificial intelligence,computers,sensors and other technologies in recent years,humanoid robots have been more and more widely used in both real life and industrial production.The improvement of robot intelligence also makes the process of human-computer interaction more comfortable.In this paper,Kinect 2.0depth camera sensor is used as a video acquisition device,and NAO robot is used as an experimental platform to conduct in-depth research on kinematics modeling and gesture recognition in the motion imitation problem.The main research contents in this paper are as follows:First of all,the research background and significance of the subject are briefly described,the development of humanoid robots and motion simulation problems at home and abroad as well as the current research status are compared and analyzed.The coordinate conversion between NAO robot and Kinect 2.0 depth somatosensory sensor is completed,the acquisition principle of Kinect 2.0 sensor depth image is introduced and compared with Kinect 1.0 sensor,and finally the processing of teaching data is introduced to lay a foundation for subsequent research.Secondly,this paper analyzes the reason for the large deviation when creating the kinematics model of humanoid robot with classical D-H model when two adjacent joints are parallel.In order to solve this problem,MD-H model,an improved model of D-H model,is used to accurately create the kinematics model of humanoid robot.Forward kinematics and inverse kinematics of the corresponding model were deduced.The effectiveness of the improved method is verified by experiments.Finally,the motion simulation system usually ignores the robot’s imitation of the hand of the teacher.To solve this problem,this paper proposes a gesture recognition method based on depth information image and integrates it into the robot’s upper limb motion simulation.The gesture recognition method proposed in this paper firstly uses the best threshold method to segment the hand region,then uses the eight-neighborhood search algorithm to extract the hand contour,and finally uses the improved center of gravity distance method combined with the depth information obtained by the visual sensor for fingertip recognition.Through comparative experiments,it is verified that the improved method effectively improves the recognition accuracy of gesture recognition.