在 Python 中使用 OpenCV triangulatePoints 时遇到问题

如何解决在 Python 中使用 OpenCV triangulatePoints 时遇到问题

亲爱的 StackOverflow 社区，我是计算机视觉的新手，正在尝试通过立体相机设置进行 3d 重建。我使用的是 OpenCV 4.2。

为了获得 3d 点，我在 Python 中使用了以下函数：

    def listener_callback(self,left_sub,right_sub):
       left_pointcloud_array=self.pointcloud2_to_array(left_sub)
       right_pointcloud_array=self.pointcloud2_to_array(right_sub)
       left_pointcloud_points=self.get_xyz_points(left_pointcloud_array,True)
       right_pointcloud_points=self.get_xyz_points(right_pointcloud_array,True)
    

       expanded_pointcloud_left=np.expand_dims(left_pointcloud_points[:,0:2],axis=0)
       expanded_pointcloud_right=np.expand_dims(right_pointcloud_points[:,axis=0)
    
       undistorted_left_points=cv2.undistortPoints(expanded_pointcloud_left,self.mtx_left,self.dist_left,R=self.R1,P=self.P1)
       undistorted_right_points=cv2.undistortPoints(expanded_pointcloud_right,self.mtx_right,self.dist_right,R=self.R2,P=self.P2)

    
    
       homogeneous_points=cv2.triangulatePoints(self.P1,self.P2,undistorted_left_points,undistorted_right_points)
    
       euclidian_points=cv2.convertPointsFromHomogeneous(homogeneous_points.T)

我使用stereoCalibrate() 和stereoRectify() 执行了校准相机的常规程序，以获得必要的参数。

我的投影和旋转矩阵如下所示：

Rotation Left:
[[ 1.74396716e-01 -2.24232058e-02 -9.84420126e-01]
 [-6.74380962e-06  9.99740653e-01 -2.27733728e-02]
 [ 9.84675472e-01  3.97824016e-03  1.74351335e-01]]
Rotation Right:
[[ 0.26564842  0.01475253 -0.9639571 ]
 [ 0.00640275  0.99983386  0.01706607]
 [ 0.96404871 -0.01070555  0.26550983]]
Projection Left:
[[802.76027664   0.         286.89848328   0.        ]
 [  0.         802.76027664 545.2985878    0.        ]
 [  0.           0.           1.           0.        ]]
Projection Right:
[[ 8.02760277e+02  0.00000000e+00  2.86898483e+02 -1.97666622e+04]
 [ 0.00000000e+00  8.02760277e+02  5.45298588e+02  0.00000000e+00]
 [ 0.00000000e+00  0.00000000e+00  1.00000000e+00  0.00000000e+00]]

每个处理步骤的输出格式如下：

Left Point Cloud Points:
[[1177.  632.    0.]
 [ 577.  645.    0.]
 [ 709.  617.    0.]
 [ 752.  599.    0.]
 [1008.  606.    0.]
 [ 933.  590.    0.]
 [ 949.  593.    0.]]
Right Pointcloud Points:
[[ 516.  617.    0.]
 [1108.  611.    0.]
 [ 899.  567.    0.]
 [ 958.  583.    0.]
 [ 716.  575.    0.]
 [ 665.  590.    0.]
 [ 875.  561.    0.]]
Left undistorted Points
[[[  -621.96014059    655.29581241]]

 [[  6237.09222297   -147.54603359]]

 [[-13996.99448978   1655.35728924]]

 [[ -6375.59001936    888.5805037 ]]

 [[ -1195.74790188    645.80792001]]

 [[ -1673.46875581    624.31269017]]

 [[ -1550.52589       628.6326525 ]]]
Right undistorted Points
[[[21831.58774037 -1164.08348054]]

 [[ -917.20195746   655.39724904]]

 [[-1774.77772702   611.04975486]]

 [[-1447.74110015   635.05964827]]

 [[-4224.1926576    715.17949778]]

 [[-6295.03254479   894.49244928]]

 [[-1940.52139146   600.20176599]]]
Homogeneous Points
[[-0.71403187  0.98887098  0.99699365  0.99187257  0.87442586  0.91830464
   0.91580335]
 [-0.44568786 -0.04820826 -0.04098027 -0.03222296 -0.07970141 -0.09989733
  -0.03444816]
 [ 0.35017816  0.1322324  -0.05588276 -0.11942512 -0.47305239 -0.3730683
  -0.40006913]
 [-0.41096969  0.04821559  0.03466128  0.02980154 -0.07244673 -0.08693191
  -0.00788962]]
Euclidian Points
[[[   1.73743195    1.08447866   -0.85207784]]

 [[  20.50936148   -0.99984793    2.74252362]]

 [[  28.76390524   -1.18230701   -1.61225349]]

 [[  33.28259031   -1.08125129   -4.00734679]]

 [[ -12.0699145     1.10013804    6.52965812]]

 [[ -10.56349263    1.14914448    4.29149981]]

 [[-116.07702337    4.36626481   50.7083032 ]]]

然而，triangulatePoints() 函数给了我完整的废话答案。我不确定我是否错误地使用了 triangulatePoints() 函数或我的相机参数是垃圾。

有人能帮我解惑吗？

非常感谢！