基于分支定界法的运动控制相机轨迹跟踪

Branch-and-bound Algorithm for Motion Control Based on Camera Tracking

提出一种与视频图像纹理信息无关,以运动控制机器人手眼标定算法为核心的相机轨迹跟踪方法。针对无特定参照物的手眼标定算法精度较低的问题,该方法提出一种运用无穷范数在给定旋转分量时迭代求局部优化解的算法。并进一步基于旋转空间分支定界法,提出新的标定方程定界约束条件及划分策略,用于寻找全局最优解。实验结果表明,提出的方法有效提高了无参照物场景的手眼标定精度,并应用于虚拟现实领域。

A video texture independent camera tracking method was proposed, which used hand eye calibration technology of motion controlfield. To resolve the problem of hand eye calibration which had less accurate result when it did not use calibration pattern, a L-infinite norm method was proposed to get local optimal solution with a given rotation. Based on a rotation branch and bound algorithm, a new bound function and branch method was proposed to get global optimal solution. Comparing with existing method, the method shows the validity in no calibration pattern sense, and it is applied to virtual reality field.