基于二阶锥规划的运动控制相机轨迹跟踪研究

Motion control camera tracking based on second cone programming

提出了一种运动控制相机轨迹跟踪方法,以解决传统基于图像特征匹配方法难以处理的存在稀疏纹理和动态物体视频图像的问题。针对运动控制中现有无特定参照物手眼标定算法精度较低的问题,对螺旋运动理论进行扩展,用于降低输入相机参数中的噪声水平,并提出了一种矩阵张量乘积的手眼标定方程表现形式,运用二阶锥规划和松弛变量约束对其优化。模拟实验和真实场景实验表明,与现有方法相比,这种方法有效提高了无特定参照物模式的手眼标定精度及其对应的运动控制相机轨迹跟踪精度。视频图像虚实融合效果的对比验证了该方法的有效性。

A motion control based camera tracking method is proposed to solve the problem of existence of sparse textures and dynamic object images,which is hard to deal with when using a conventional image feature detector based camera tracking method.To resolve the problem of hand eye calibration in motion control,which has less accurate result when it does not use calibration pattern,a screw motion based camera data selection strategy is proposed to reduce the input camera motions noise.The hand eye calibration function is changed by kronecker product and solved by second cone programming and slack variances constrain methods.Comparing with existing methods in a simulated and a real environment,the proposed method shows the validity in no calibration pattern sense,and it is applied to virtual reality field.