Please select Into the mobile phone version | Continue to access the computer ver.

关于3D bounding box计算问题

[Copy Link]
Author: qulei | Time: 2022-2-22 14:26:16 | Others|
2 3444

4

Threads

9

Posts

48

Credits

Kindergarten

Rank: 1

Credits
48
Posted on 2022-2-22 14:26:16| All floors |Read mode


最近在学习 livox_detection 源码,运行测试效果很好, 关于3D bounding box计算有一些疑问??
  1. T1 = np.array([[0.0, -1.0, 0.0, 0.0],
  2.                [0.0, 0.0, -1.0, 0.0],
  3.                [1.0, 0.0, 0.0, 0.0],
  4.                [0.0, 0.0, 0.0, 1.0]]
  5.               )
  6. box3d_pts_3d[:, 0:3] = self.get_3d_box((l, w, h), ry,(x,y,z))
  7. box3d_pts_3d = np.dot(np.linalg.inv(T1), box3d_pts_3d.T).T
Copy the code
根据get_3d_box()函数的计算得到检测框的8个角点坐标。
为什么需要使用 T1 矩阵进行变换才能得到正确的角点坐标?
这个 T1 矩阵是什么意思?
(x,y,z) 不是参考雷达坐标系吗,所以才进行一次变换?



Reply

Use props Report

2

Threads

267

Posts

1079

Credits

Administrator

Rank: 9Rank: 9Rank: 9

Credits
1079
Posted on 2022-2-22 15:59:49| All floors
一个是相机坐标系,一个是雷达坐标系,T1实现坐标系的转换
Reply

Use props Report

4

Threads

9

Posts

48

Credits

Kindergarten

Rank: 1

Credits
48
 Author| Posted on 2022-3-1 15:25:15| All floors
Livox Support Posted at 2022-2-22 15:59
一个是相机坐标系,一个是雷达坐标系,T1实现坐标系的转换

根据雷达的点云的检测,怎么会出现相机的坐标系呢?
相机的位置也不会和雷达在同一个点,T1这个坐标转化,只是重新调整的xyz的顺序?
Reply

Use props Report

You need to log in before you can reply Login | Register

Credit Rules

Quick Reply Back to top