| HCI is an effective way which can realize the interaction between people and computers through input and output devices.The rapid development of technology in the field makes the human-computer interaction more comfortable and more natural.And it came into a variety of human-computer interaction models, such as data, touch devices, sonic, human action modes.This paper involves projection-camera system which is the use of overhead projector, camcorder as input and output devices.And it combined with human action as a interactive mode in man-machine interactive systems.It is a new multimedia display interactive platform,which uses display technology, computer vision, pattern recognition and artificial intelligence technology to create an interactive fantasy dynamic experience.There exists the following problems. Firstly, it exists large deformation. Secondly, It is difficult to extract feature gesture. Finally, a large amount of computation can not meet the requirement of real-time system.This paper mainly for the above problems,transfer the habits of which people controls the touch screen to the projector-camera system. Put forward to use gesture shadow as communication controller because of interactive gesture block the light of projector.Firstly, research and analysis the photometric model of camera.Consider the effects on the accuracy of the image because of the one order radial distortion and tangential distortion of camera.Calibrate the camera and obtain the internal and external parameters of camera.Combined with the principle of camera distortion, use affine transformation algorithm to do geometric mapping for projector and camera system.Then, use Gauss mixture model of hybrid position the moving objects.Use Otsu to detect the shadow gestures because interacting persons stand in front of projector which block the light of projector. The combination of color mixture model to detect the color of gestures, and according to the real interaction gesture must be around the shadow, track the interaction gesture. Then, according to the finger tip with strong symmetrical features,use radial symmetry transform to locate the primary position.The semicircle of fingertip intersect at two points with fingers.And the fingertip and two points construct two lines. The angle of two lines has a strong constraint angle at the tip of the finger.So it can position the fingertip. At the same time, according to the characteristics of action of Double-click,use DTW to distinguish the gesture.Finally, build platform to realize the projector-camera interactive system which based on SOPC experimental box.Its"main processor is Virtex-4FX which is developed from Xilinx company.Use the way of hardware to get the video from camera,and decode the data of video based on the the PAL format.Then use the ping-pong algorithm to reduce the consumption of the internal storage resource, and improve the video acquisition rate.And construct the IP kernel of DTW algorithm to improve the real-time performance of system. |
