将图像坐标一点转化为三维物理坐标Python代码

要将图像坐标转化为三维物理坐标，需要知道图像的相机内参和相机外参，可以使用以下代码实现：

import numpy as np

# define camera intrinsic matrix
K = np.array([[fx, 0, cx], [0, fy, cy], [0, 0, 1]])

# define camera extrinsic matrix
R = np.array([[r11, r12, r13], [r21, r22, r23], [r31, r32, r33]])
t = np.array([tx, ty, tz])
T = np.concatenate((R, t.reshape(3, 1)), axis=1)
P = K @ T

# image point
u = 100
v = 200

# homogeneous image point
uv = np.array([u, v, 1])

# compute ray direction in camera coordinate
ray_dir = np.linalg.inv(K) @ uv

# compute intersection point with ground plane
ground_z = 0
ray_origin = -R.T @ t
t = (ground_z - ray_origin[2]) / ray_dir[2]
intersection = ray_origin + t * ray_dir

# convert to world coordinate
world_point = np.linalg.inv(R) @ (intersection - t)

其中，`fx`、`fy`、`cx`、`cy`分别是相机的焦距和主点坐标，`r11`~`r33`和`tx`、`ty`、`tz`是相机的旋转和平移参数。`u`和`v`是图像中的像素坐标，`uv`是齐次坐标。代码中先计算光线在相机坐标系中的方向，然后计算光线与地面平面的交点，最后将交点从相机坐标系转换到世界坐标系。