基于对称重复纹理的摄像机位姿测量方法

Camera pose determination method based on repetitive and symmetrical texture

某些目标上具有对称重复的纹理,这些纹理的图像矩阵具有低秩特性。当摄像机在偏离对正方向拍摄纹理时,由于存在投影形变,该纹理图像在原有方向上的对称性和重复性不再保留,从而不再具备低秩特性。通过建立摄像机拍摄方向与纹理图像的秩之间的数学模型,将摄像机方向的求解问题等效为一定约束下纹理图像的秩最小化问题。求解该问题,可得到摄像机拍摄方向,进一步结合纹理上任意两点的成像关系求解摄像机的空间位置。该方法与传统的基于目标图像中点、线等局部特征的测量方法相比,具有更高的准确性和良好的抗噪性能。仿真实验和实际实验均证明了该方法的有效性。

Repetitive and symmetrical texture exists on many targets,and the images of these targets can always be modeled as ＂lowrank＂matrices. When the camera shoots the target at tilted position,the symmetry and repetition of the texture image existing in the original direction cannot be retained and the image matrix no longer has＂low-rank＂characteristic due to the projective deformation. By establishing the mathematic model between the camera pose and the rank of the texture image the problem of solving the camera orientation is made equivalent to a rank minimization problem of the texture image under certain constraints. Through solving this problem,the camera shooting direction can be obtained,and the camera spatial position can be calculated with the imaging relation of arbitrary two points on the texture. Compared with the traditional local characteristic pose determination methods based on the points or lines in the target images,the proposed method possesses better accuracy and noise resistance capability. Both simulation and actual experiments were conducted,and the results show the effectiveness of the proposed method.