Research On Mobile Robot Controlled By Static And Dynamic Gesture

Robot control is achieved generally in advance to input the control instructions to the computer hardware, which is lack of interaction between people and the robot. Hand gesture is a way to control robot, which is converted to a mobile robot control commands and can remotely control the robot. Hand gestures are divided into static and dynamic two, the dynamic hand gesture is a certain time series within the hand movements, and static hand gesture is for a moment. In recent years, gesture recognition technology is usually using the camera to obtain hand images to identify, Microsoft launched Kinect camera has good performance in the gesture recognition, which access to deep image information and human skeletal data.In this paper, Kinect camera is used to realize static and dynamic hand gesture recognition, and then the result is applied to the control of mobile robot.The main contents are as follows:(1) In this paper, a fingertip detection algorithm based on convex hull analysis and curvature is proposed. Firstly, the hand region is divided by the double depth threshold, and the noise in the image is processed by median filtering and morphological operation. According to the contours of the hand to find the hand convex hull, and finally use the convex hull information and curvature features to select the fingertips.(2) In this paper, a static gesture recognition algorithm based on SVM classifier is proposed, the feature vector includes the number of fingertips, the angles between fingertips, the distances between fingertips and palm, and the horizontal angle of fingertips and hand. Realize static gesture recognition with SVM, because its outstanding performance on non-linear data samples and small sample data samples.(3) In this paper, we propose a weighted DTW dynamic gesture recognition algorithm. Firstly, we extract the displacement of six joints as the features of dynamic gestures by Kinect. Considering that there are not many displacements in the dynamic gestures, the joint weight is calculated according to the displacement in each dynamic gesture, and the dynamic gesture similarity is recognized by the DTW algorithm.(4) The aim of hand gesture recognition is used for mobile robot gesture control, this paper uses an intelligent car as mobile robot, send the command which is corresponding to the hand gesture with wireless. Smart car uses Arduino as a controller, which drives the ultrasonic module rotation with a steering gear to detect the surrounding obstacles.The experimental results show that the static and dynamic gesture recognition algorithms can meet the requirement of mobile robot control. And depth images acquired by Kinect camera can avoid the influence of complex environment and light, which has a good research value.