Research On Gesture Interaction Control Technology For Mobile Robot

In recent years,robots have been a hot area of research in many countries,which are widely used in security,entertainment,education,industry and other fields.The Human-Computer Interaction(HCI)is also constantly updated.Since Microsoft launched Kinect somatosensory components in 2010,non-contact human-computer interaction technology has gradually emerged and widely used in a variety of robot control fields.By using Kinect2.0,a mobile robot based on gesture control is designed in this research which focuses on gesture interaction control technology including gesture definitions,gesture acquisition,gesture preprocessing and gesture priority recognition.This robot can make corresponding action by recognizing different gestures,and can complete the function of omnibearing moving in the room,avoiding obstacles and detecting the movement state of robot,which is laying the foundation for the design and realization of indoor service robot.The robot can be used to help the daily life of elder and disabled people and improve their quality of life in the future.It can also be used in the field of business services.The whole robot system is divided into two parts: slave computer and host computer.The slave computer uses STM32F103ZET6 as main controller,with which external connects six ultrasonic sensors and one MPU9250 sensor are connected outside.The main function of the slave computer is to receive the command form the host computer and complete the corresponding action accordingly.The slave computer also has the ability to avoid obstacles automatically and detect the movement states of itself.The host computer is composed of i7 dual-core industrial computer installed with Windows8.1 operation system and Kinect2.0,which is mainly used to complete the acquisition of the video signal,gesture recognition and processing,and form the command sent to the slave computer.It can also read the status of the slave computer.The host computer and the slave computer exchange information through RS232 interface.Kinect2.0 can obtain color image data,depth image data and human skeletal data.The depth image acquired has good robustness since it is not affected by light and texture.The human skeletal data includes 25 skeletal nodes.When a large area of human body is under the vision of the robot,it has the advantages of locking the skeletal position accurately,obtaining the skeleton node fast,tracking the effect well and so on.Therefore,this research focuses on the method of gesture recognition based on the depth image.In this method,firstly,the hand image is quickly locked by the skeletal data.Secondly,a image difference method based on a double threshold(one is the fixed threshold and one is the dynamic threshold)for separating the background and hand is proposed to extract the binary hand image.Finally,gesture based on finger's numbers is recognized by using gesture point counting method.Based on the Visual Studio 2012 development environment,the open source computer vision library(OpenCV3.0)and the Kinect2.0 class library,the host computer software is developed by using C++ language.And the slave computer software is designed in the Keil4 development environment.After testing,the entire system can accurately identify six kinds of right hand gestures,and the robot can make appropriate action according to the gesture command.Test results show that the average recognition rate of gesture is 92.5%.The system can also recognize the left hand gestures.Thus,by combining the left and right hands gestures,more gestures may be formed for HCI.