decomposeEssentialMat() always return normalized T (Bug #3423)

Added by Yoshinari Kameda over 11 years ago. Updated about 11 years ago.

Status:Cancelled Start date:2013-12-06
Priority:Normal Due date:
Assignee:Yoshinari Kameda % Done:

0%
Category:calibration, 3d
Target version:3.0
Affected version:branch 'master' (3.0-dev) Operating System:Linux
Difficulty: HW Platform:x86
Pull request:

Description

Hi,

It seems decomposeEssentialMat() always returns normalized T (|T|=1.0).
T should be a transition vector, so it should have some proper length, right?
I have checked it only at python.
(It happens at recoverPose() too.)

R0,R1,T = cv2.decomposeEssentialMat(E)

R0,R1 seems to be fine.
Attached is my quick code to check that. Ple

---FYI---
I got the code on 2013/11/15 by
git clone git://github.com/Itseez/opencv.git

recoverPose_debug00.py - script to check the function behavior (1.3 kB) Yoshinari Kameda, 2013-12-06 06:24 pm

recoverPose_debug00.txt - output of the script (669 Bytes) Yoshinari Kameda, 2013-12-06 06:24 pm

History

#1
Updated by Nikita Manovich about 11 years ago

Hi,

Thank you for the bug report. Could you please help us to investigate and fix the problem? If you can prepare a pull request to fix the problem it will help us a lot. Please look at the article: http://www.code.opencv.org/projects/opencv/wiki/How_to_contribute

  • Assignee set to Yoshinari Kameda
  • Status changed from New to Open

#2
Updated by Bo Li about 11 years ago

Hi this is not a problem. By decomposing essential matrix, you are not able to recover the full translation but only the direction. I've sent a pr to update the doc. For the code attached: your generated E is scaled by the length of t. However, in practice when you estimate an essential or fundamental matrix, you don't have scale.

  • Status changed from Open to Cancelled

Also available in: Atom PDF