0

I have some consecutive images,and had extract the keypoints of every image.

If these images all has 8 same keypoints,is there any algorithm to build the 8

keypoints' world coordinate?

I had googled that euclidean reconstruction may could done that,is it right?

Thank you for your help!

zdczdcc
  • 119
  • 2
  • 14
  • Look up Structure From Motion – rayryeng Jul 10 '16 at 05:59
  • @rayryeng,ok,thank you! – zdczdcc Jul 12 '16 at 03:02
  • @rayryeng i have implemented the linear ls triangulation.But i met a new problem. The triangulation between 2 imgs,such as `img1` and `img2`.`p1` and `p2` is the projection of 2 imgs.In this situation,the algorithm consider the world coordinate system is `img1`'s camera coordinate system(tranformed from real world coordinate system).But i have a marker in the 2 imgs.I want to reconstruct the keypoints in the marker's coordinate system(i have defined.The marker's shape is square.)Could you help me ,thank you! – zdczdcc Jul 14 '16 at 10:39
  • @rayryeng Now my idea is calculatting `p1` from `img1`,`p2` from `img2`(Get from the marker.Detected the marker's pixel coordinate and defined the marker's real world coordinate.Then slovePnP to get the rotation_mat and tran_mat).Input `p1`,`p2` and matched_keypoints in 2 imgs to the svd to calculate the each matched_keypoints' real world coordinate(original is the marker).But I got error result.x,y,z were all strange.I don't know why. – zdczdcc Jul 14 '16 at 11:14

0 Answers0