Research On Remote Visual Perception And Presentation Methods

Teleoperation using unmanned machinery is an effective technology for working in environments which are generally inaccessible or hazardous to humans.The traditional unmanned machinery relies on stationary camera system for remote scenes shooting and uses typical screen-based display for video presentation.However,this kind of method is largely limited by the narrow angle of view,the disparity between camera adjustment and visual habits of users,the lack of depth perception about the remote site.As a result,the efficiency of teleoperation will be reduced,meanwhile task error rates will increase.Therefore,it is of significant importance to study and develop a remote visual perception system which can provide stereo perception,immersion,and wide field-of-view(FOV).To address the problems mentioned above,the methods of remote visual perception and presentation are studied and explored.The main tasks and innovations include:(1)A remote visual perception experiment system is set up.The system takes advantage of a binocular camera to capture stereoscopic video in remote scenes,a head follow-up controlled high-speed Pan/Tilt to increase the FOV,as well as a head-mounted display to enhance the stereo feeling and immersion.The system is easy to construct and fully functional,and can provide users with an experience of telepresence.(2)The method of processing and rendering stereo video is improved.First of all,the vertical parallax is eliminated through performing geometrical correction based on a simple and effective rotation transformation method according to the principle of binocular parallax fusion.Then the differences of brightness which can affect viewing comfortableness are reduced by performing histogram matching.Finally,a scale transformation method based on the displaying model of HMD is proposed to improve the consistency of depth perception of observation by HMD compared to naked eyes.(3)The Augmented Reality(AR)in stereo vision is realized.The speed of registration of AR in stereo vision is accelerated by marker detection in downsampled images and retrieve of pixel coordinate points.This method is applied to the system test,and the stereoscopic acuity of the system is measured by the improved Howard-Dolman test method.The egocentric distance perception provided by the system is tested by verbal estimation,and the effectiveness of the scale transformation method is also verified.