I am wondering what the best way to use a digital camera to measure
the spatial distance (in the real world) between two points A an B
is. Assume that the points A and B are on a plane P that is placed
orthogonally to the camera. I already know the distance from the
camera and the plane P.
I envision a simple method in which I measure two arbitrary points C
and D on the plane P using a ruler. Then I use the camera to take a
picture of the plane P. Let's denote the images of the two points C
and D on the captured image as C' and D'. I compute the ratio CD/C'D'.
Now, to measure the distance between A and B on the plane P, I just
compute A'B' * CD/C'D', where A'B' is the distance in pixels between
the images A and B' of A and B, respectively on the captured image.
I guess since the plane P is orthogonal to the camera len, this method
may work. But of course, I am not sure.
So computer vision experts, please help.
Thanks
Even assuming that your lens does not introduce any distortion in the
image (pincushion, barrel, etc, which is a bit of a dangerous
assumption for many small digital cameras, your method is inadequate.
If you will make a drawing looking down on the camera and the
distances AB and CD you will see that one of them is farther from the
camera than the other. Consequently the simple ratio does not work -
you would also have to correct for the ratio of distances. Single-view
metrology, in which an image from a single viewpoint is used to make
measurements, requires knowing quite a lot about the geometry of the
scene and the response of the camera. If you are serious about it,
read a book like K. B.Atkinson "Close Range Photogrammetry and Machine
Vision" Whittles Publishing, 2001.
Yes, I am very serious about this issue. And definetely I will read
the book.
If you know any other books, references, please let me know.
Best,
> Vision" Whittles Publishing, 2001.- Hide quoted text -
>
> - Show quoted text -