Research On Gesture Control Technology Based On Interaction With Robot

With the fast development of image recognition technology and robot control technology, Now master-slave control technology of visual robot based on the gesture recognition more and more attract people’s attention. It is quite necessary to find a way out to figure out the problem about How to control robot to finish the specific task, and how to transmission the information to robot succinctly and high efficiency without Bluetooth or infrared or wearable devices. Because gesture control has the advantage of parallelism which means one gesture can convey many information in the short time. It will also mitigate the communication equipment. So gestures control visual robot has great practical significance.Usually gesture images trend to be translation scaling rotation or shearing and so on posed by the different people or different time. This thesis focuses on How to find out the gesture image under the complex circumstances, and extract the gesture information, classify the gesture according to the information.Geometry feature based hand recognition algorithm is quite fast, however, it is suit for specific gestures. Polar radius moment model has the translation invariance and rotation invariance. Polar radius of central moment has the translation invariance, rotation invariance and scale invariance. So it is suitable for the gesture image. However, it is not suitable for the none-scale transformation. So this thesis combine these two different model to make recognize gesture image fast.First find out the gesture outline according color distribution in the HSV color model. Then find out the centroid of the gesture and mark off the finger area to identify roughly the class which the gesture belongs to. At the same time classify the polar radius of central moment information. Combine with this two steps to classify the gesture accurately.Finally using above model do the experiment on the fully autonomous robot platform. Test on different gesture posed by different people, at different time and at different complex circumstance. Experiment result shows that it classify the gesture accurately based on the marking off the finger area combined with the polar radius of central moment of the gesture outline. At the same time it can handle image in the real time to satisfy the robot control realtime requisition.