Research On Kinect Sensor-based Motion Control Methods For Humanoid Robot Hands

With the development of science and technology,human-computer interaction gradually return to the most natural way,such as through gestures,voice,eyes,facial expressions and other ways as the computer,mobile device input,these ways are simple and convenient,and do not need additional input devices,such as keyboard and mouse.Gesture-based human-computer interaction is a popular research area,and it plays an important role in the human-computer interaction.In recent years,the rapid development of industrial robots,resulting in many of workers in the traditional factories are gradually replaced by the industrial robots.At present,most of the robots are based on the traditional control methods,the control processes are complex and often need the relevant professional knowledge,and sometimes have to remember a lot of operating commands.Whether the service robots or the industrial robots,the operation of their end-effectors are moving in a more complex,flexible,versatile direction of development.In fact,the humanoid robot hands are the future direction of development,and the humanoid robot hands are more freedom,high flexibility,complex control,If the use of the traditional control methods will be very troublesome,especially in the teaching process,therefore,the study of the gesture-based humanoid robot hand movement control is of great significance.In this thesis,a method of Kinect sensor-based motion control for the humanoid robot hands is proposed.The kinematics model of the humanoid robot hand is established by using D-H method and the positive kinematics equation,the inverse kinematics equation as well as the static equation are all deduced in detail.The 3D model of the humanoid robot hand was established by the 3dsMax software,and the correctness of the kinematics model was verified by simulations.The grasping planning is one of the most important parts of the humanoid robot hand motion control,this thesis mainly also studies the simple shape grasp planning,which includes the mathematical modeling of the object to be grasped and the contact point position planning of the balanced grasping.Based on the research of the dual threshold hand gesture segmentation method which is based on depth image and the ellipse model based on skin color,a new segmentation algorithm based on the double threshold method,skin color detection and watershed algorithm is proposed,by using the Kinect sensor to obtain the depth data,skeleton information,color images to complete the gesture segmentation.Using the different algorithms to experiments,it is known that the proposed algorithm has good effect and can meet the real-time requirements.This thesis also used the optimized DBN model to complete the gesture recognition,and achieved good results.Finally,the design of the motion control system of the humanoid robot hand based on the Kinect sensor is introduced,through the proposed system can complete the simulation and control the movement of the humanoid robot hand,and it also supports the control of the two(left and right)humanoid robot hands.