CAJ | 학술논문

为了提高遥操作的操作效率，采用一种基于单目视觉的预测显示方法来解决时延导致的视觉反馈滞后问题．该方法通过基于地图的相机位姿估计算法来实时跟踪机器人的状态，在线构建机器人工作环境的三维几何结构模型，并结合多纹理映射技术进行渲染，将模型重投影到预测视点下，得到逼真的预测图像．搭建了一个基于客户端服务器端模式的系统平台，用于未知环境下的遥操作．结果表明，在总长为7．112 m 的摄像机运动轨迹中，位姿跟踪的平均误差约为0．015 m．该系统不仅能提供预测图像，而且支持生成任意视点的图像，有利于操作者从各个角度观察机器人工作场景．
위료제고요조작적조작효솔，채용일충기우단목시각적예측현시방법래해결시연도치적시각반궤체후문제．해방법통과기우지도적상궤위자고계산법래실시근종궤기인적상태，재선구건궤기인공작배경적삼유궤하결구모형，병결합다문리영사기술진행선염，장모형중투영도예측시점하，득도핍진적예측도상．탑건료일개기우객호단복무기단모식적계통평태，용우미지배경하적요조작．결과표명，재총장위7．112 m 적섭상궤운동궤적중，위자근종적평균오차약위0．015 m．해계통불부능제공예측도상，이차지지생성임의시점적도상，유리우조작자종각개각도관찰궤기인공작장경．
To improve the operation efficiency of teleoperation, a predictive display method based on monocular vision is proposed for solving the visual feedback delay problem caused by time delay.In this method, the robot's poses are tracked in real-time by the map-based camera pose estimation al-gorithm.The three-dimensional geometry model under robot working environments is reconstructed on-line with rendering by multiple texture mapping technology.Finally, the realistic predicted image is obtained by projecting the model into the predicted view.The system platform based on the client-server architecture is constructed which is suitable for teleoperation under unknown environments. The experimental results show that the average error of pose tracking is about 0.015 m over a camera journey of 7.112 m.The proposed system can not only supply the predicted images, but also sup-port to generate the images of arbitrary views, which benefits the operator observing the working en-vironments of the robot from all angles.