基于矩形两组对边的消隐点特性和隐含的长宽比信息，提出了一种新的摄像机自标定几何方法。该方法仅依据同一个矩形的两次或三次成像，即可在摄像机传感器特性已知或未知时标定摄像机内参数并辨识矩形长宽比。利用空间中有限距离点与同一无穷远点的连线相互平行和完全四边形的调和分割特性，以及被多次成像的矩形长宽比相同的特点，建立了摄像机内参数约束方程。通过建立与直线段成像相关的代价函数，提出了畸变参数寻优与线性内参数标定相迭代的畸变校正方法，可获得与摄像机无畸变情况下相当的自标定精度。在确定矩形任意两个顶点坐标的情况下，即可求解摄像机所有外参数。仿真实验表明，该标定算法收敛快速，对图像噪声不敏感。实际图像实验表明，与传统平面靶标法相比，该方法不但减少了预知条件，而且提高了标定精度和效率。

A new camera self-calibration approach is proposed based on the property of the vanishing points and the aspect ratio of the rectangle′s two groups of opposite sides. This method can calibrate the camera′s intrinsic parameters and identify the aspect ratio of the rectangle, whether known or unknown the character of camera′s sensor, according to twice or triple imaging for the same rectangle, respectively. The equations of the camera′s intrinsic parameters are established via three properties: the first one is that the lines which connect finite points and the same infinite points are parallel; the second one is the harmonic division which consists in the complete quadrilateral; the third one is the identity length-width ratio of the rectangle which imaged sometimes. A correction method of the camera distortion, utilizing which the accuracy of self-calibration correspond to the no distortion situation, is proposed via iterative between the optimization of nonlinear distortion parameters and solving of linear intrinsic parameters, based on constructing a cost function of lines′ imaging. Simulations prove that the calibration algorithm can converge sharply, and the results are not sensitive to image noise. Real imaging tests prove that, comparing with traditional calibration by flat surface drone, this method can reduce foreknowledge conditions, as well as promoting both precision and efficiency of the calibration results.