Halcon - 相当于 affine_trans_contour_xld 的 3D
Halcon - 3D equivalent of affine_trans_contour_xld
我有一个 XLD 轮廓，我需要将其从世界平面转换为图像平面..但我只能找到 affine_trans_contour_xld，它接受 2D 垫，而不是 3D 垫..我怎么能将 xld 轮廓从世界平面转换为图像平面？
基本上我需要的是contour_to_world_plane_xld
的倒数
编辑：我认为解决方案是找到 XY 平面的 2D 垫，但也找不到如何做到这一点..
我写了一段我以前用过的代码。
程序名称：world_plane_coordinates_to_image_points
输入控制参数：
CamParam、姿势、X、Y
输出控制参数：
行、列
* given the X, Y coordinates on the world plane,
* the procedure returns the relative points on the image (original or unwarped)
* create arrays with z coordinates all equal to 0
tuple_length (X, NumPoints)
tuple_gen_const (NumPoints, 0, Z)
* Coordinates in the distorted image
pose_to_hom_mat3d(Pose, HomMat3D)
affine_trans_point_3d(HomMat3D, X, Y, Z, Qx, Qy, Qz)
project_3d_point(Qx, Qy, Qz, CamParam, Row, Column)
您可以通过阅读文档更好地理解解决方案：
https://www.mvtec.com/fileadmin/Redaktion/mvtec.com/documentation/solution_guide/solution_guide_iii_c_3d_vision.pdf
信息位于：
3.3.5 将世界坐标转换为图像坐标
我是这样解决的：
*get the points of a 100 mm square on my XY plane. Since the XLD is in 1px=1mm
affine_trans_point_3d(Mat3D, 0, 0, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp1, Yp1)
affine_trans_point_3d(Mat3D, 0, 0.1, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp2, Yp2)
affine_trans_point_3d(Mat3D, 0.1, 0.1, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp3, Yp3)
affine_trans_point_3d(Mat3D, 0.1, 0, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp4, Yp4)
*0 on my Image corrisponds to 0 on my XY plane. 100 px square equals the 0.1m square above.
*generate the 2D mat by projection
hom_vector_to_proj_hom_mat2d ([0,100,100,0], [0,0,100,100], [1,1,1,1],
[Xp1,Xp2,Xp3,Xp4], [Yp1,Yp2,Yp3,Yp4], [1,1,1,1], 'dlt', HomMat2D)
*load the first Z plane
select_obj(ListZplanes, CurrentZplaneXLD,i)
*project the Z plane onto the Plane
projective_trans_contour_xld(CurrentZplaneXLD,CurrentZplaneTransXLD,HomMat2D)
我有一个 XLD 轮廓，我需要将其从世界平面转换为图像平面..但我只能找到 affine_trans_contour_xld，它接受 2D 垫，而不是 3D 垫..我怎么能将 xld 轮廓从世界平面转换为图像平面？
基本上我需要的是contour_to_world_plane_xld
编辑：我认为解决方案是找到 XY 平面的 2D 垫，但也找不到如何做到这一点..
我写了一段我以前用过的代码。
程序名称：world_plane_coordinates_to_image_points
输入控制参数： CamParam、姿势、X、Y
输出控制参数： 行、列
* given the X, Y coordinates on the world plane,
* the procedure returns the relative points on the image (original or unwarped)
* create arrays with z coordinates all equal to 0
tuple_length (X, NumPoints)
tuple_gen_const (NumPoints, 0, Z)
* Coordinates in the distorted image
pose_to_hom_mat3d(Pose, HomMat3D)
affine_trans_point_3d(HomMat3D, X, Y, Z, Qx, Qy, Qz)
project_3d_point(Qx, Qy, Qz, CamParam, Row, Column)
您可以通过阅读文档更好地理解解决方案： https://www.mvtec.com/fileadmin/Redaktion/mvtec.com/documentation/solution_guide/solution_guide_iii_c_3d_vision.pdf
信息位于：
3.3.5 将世界坐标转换为图像坐标
我是这样解决的：
*get the points of a 100 mm square on my XY plane. Since the XLD is in 1px=1mm
affine_trans_point_3d(Mat3D, 0, 0, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp1, Yp1)
affine_trans_point_3d(Mat3D, 0, 0.1, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp2, Yp2)
affine_trans_point_3d(Mat3D, 0.1, 0.1, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp3, Yp3)
affine_trans_point_3d(Mat3D, 0.1, 0, 0, X, Y, Z)
project_3d_point(X, Y, Z, CamParam, Xp4, Yp4)
*0 on my Image corrisponds to 0 on my XY plane. 100 px square equals the 0.1m square above.
*generate the 2D mat by projection
hom_vector_to_proj_hom_mat2d ([0,100,100,0], [0,0,100,100], [1,1,1,1],
[Xp1,Xp2,Xp3,Xp4], [Yp1,Yp2,Yp3,Yp4], [1,1,1,1], 'dlt', HomMat2D)
*load the first Z plane
select_obj(ListZplanes, CurrentZplaneXLD,i)
*project the Z plane onto the Plane
projective_trans_contour_xld(CurrentZplaneXLD,CurrentZplaneTransXLD,HomMat2D)