中文摘要：

针对现有的多摄像头全自动目标定方法要求至少3幅不同视图才能完成标定的限制条件,为减少摄像头数量、降低摆放设计复杂度以及提高系统的整体性能,通过严格的数学推论证明了在摄像头内参一致的约束下,只需要2幅不同视图即可完成标定.经实验验证,对标定结果进行二维重映射的像素误差约1个像素.该算法在更宽松的限制条件下,仍然保持了操作简便、计算结果稳定且精度较高的性能.

英文摘要：

Now existing automatic self-calibration method has a limitation of requiring not less than three different views to complete the calibration.In order to reduce the camera amount,complexity of the camera placement,and enhance the performance of the whole system,rigorous mathematical reasoning is used to prove that two different views are enough if exploiting the constraint of invariable intrinsic parameters.The experiments show that the calibration result can achieve about 1 pixel error in 2D reprojection test.This algorithm has convenient pre-process and robust,accurate results under loosened limits as well as the state-of-the-art.