【摘要】 用单摄像机所获取的二维(2D)图像来估计两坐标之间的相对位姿和运动在实际应用中是可取的,其难点是从物体的三维(3D)特征投影到2D图像特征的过程是一个非线性变换,把基于单目视觉的位姿和运动估计系统定义为一个非线性随机模型,分别以迭代扩展卡尔曼滤波器(IEKF)、一阶斯梯林插值滤波器(DD1)和二阶斯梯林插值滤波器(DD2)作非线性状态估计器来估计位姿和运动.为了验证每种估计器的相对优点,用文中所提方法对每种估计器都作了仿真实验,实验结果表明DD1和DD2滤波器的特性要比IEKF好.更多还原

【Abstract】 A solution to relative pose and motion estimation between two reference coordinates that used twodimensional (2D) intensity images from a single camera was desirable for realtime applications.The difficulty in performing this measurement was that the process of projecting threedimensional (3D) object features to 2D images was a nonlinear transformation.The system of pose and motion estimation which is based on the monocular vision was defined as a nonlinear stochastic model.The system used the iterated extended Kalman (filter) (IEKF),the firstorder Stirling’s interpolation filter (DD1) and the secondorder Stirling’s interpolation filter (DD2) respectively as nonlinear state estimators to estimate pose and motion.The method has been implemented with simulated data based on three kinds of different estimator respectively to show the relative advantages of each kind estimator,and the simulation result has shown that the performance of DD1 and DD2 is superior to IEKF.更多还原