上下文：我正在研究projection equations。
我有 N 个投影矩阵（每个2D点），形状为(3, 4)。
例如，假设P = [p_i] for 1 <= i <= n。
import numpy as np
# p_i were initialised with dummy values
p1 = np.array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]], dtype=float)
...
pn = np.array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]], dtype=float)
然后我在摄像机框架中得到了 N 个2D点坐标（形状为(3, 1)）。
# coords_i were initialised with dummy values for x, y and (z was set to 1 by design) as well
coords_1 = np.array([10, 20, 1], dtype=float)
...
coords_n = np.array([30, 40, 1], dtype=float)
例如，假设C = [coords_i] for 1 <= i <= n。
任务是找到这样的X，它是单个(4, 1)向量，以便与p_i * X = coords_i相乘。
如何将这个系统传递给numpy的linalg solver？
P * X = C
所以最终形状为P.shape = (3 * N, 4), X.shape = (4, 1), C.shape = (3 * N, 1)