针对几何参数未知的非合作目标,提出一种基于双目视觉的非合作目标相对位置和姿态估计算法.对目标图像进行滤波去噪、Canny边缘提取并将每条边缘段保存至链表,将边缘段分为直线段和圆弧段;并从直线段和圆弧段中检测出矩形特征和圆形特征;根据平行直线在同一像平面投影具有相同湮灭点的性质设计矩形特征相对位置和姿态估计算法;根据圆在像平面对偶投影几何,设计圆形特征的相对位置和姿态及半径的估计算法;对实际模型图像进行处理验证,结果表明该算法能够很好地识别帆板和对接环,并有效地估计出相对位置和姿态及对接环半径.
A relative estimation method of non-cooperative spacecraft is proposed based on stereo vision with unknown geometric parameters. The image of target is smoothed and thinned by Canny operator. Edge segments, represented by linked-lists, are classified into straight segments and arc segments. Rec- tangle features and circle features are recognized from straight and are segments respectively. Algorithm for relative position and attitude estimation of rectangle is designed based on the property of vanishing point of parallel lines. We also presented a novel and efficient method of estimating relative position and attitude and radius of circle. At the end, real images are tested to validate the proposed algorithms.