4
$\begingroup$

I have a situation where i need to extract the top view of a road from the 'skewed' view I get from a cc camera located on some roadside tree. I would like to do this using OpenCV, but since I'm very new to image processing and really don't know what approach to take, please provide me any sort of guidance on how to approach the problem.

$\endgroup$

migrated from stackoverflow.com Oct 9 '12 at 12:18

This question came from our site for professional and enthusiast programmers.

  • $\begingroup$ i'm very new to open cv/ image processing so i really don't know how to approach this in order to try something :/ $\endgroup$ – Aryan Oct 9 '12 at 6:05
  • 2
    $\begingroup$ Did you google it? Did you read the documentation/tutorials in opencv.org? Did you look for academic papers on your subject? SO is not a replacement for google $\endgroup$ – vasile Oct 9 '12 at 7:01
7
$\begingroup$

I am not giving you a complete algorithm here, but since I worked on a similar project, I can give you some hints and tips.

First of all, changing an image taken from one perspective to a different one relatively easy only for planar surfaces. E.g., you have an image of a tall building with a road, and since buildings are usually build vertical in the air and roads tend to be flat on the ground, the planes of the building and the road are perpendicular. Now, if you want to get an image where you are looking straight in to the building (e.g. building plane parallel to image plane), you have two options:

  • either transform the image so that the building plane is the way you want it, and disregard the rest of the image (road plane will not be correctly transformed)
  • in addition to transforming the building plane, calculate and perform a transformation on the image part representing the road plane, and somehow stitch them together.

The transformation that allows you to "change" the perspective is called the planar perspective transformation or homography. A homography represents a (mathematical) connection between two images of a same object taken from different perspective, e.g. allows you to calculate the unknown image coordinates of a point in space in one of the images from known image coordinates of the same point in the image with a different perspective.

To be able to calculate a homography, you need to know four pairs of corresponding points. In case of bad camera, you will also have to take care of the camera intrinsic parameters. A very good post on this can be found here.

Now, if you really want to do it properly, you should understand all the terms I used up until now really well before doing anything else. That said, I have been working on a fairly large project, and some parts required to transform images of a road taken from a moving car in to images taken from the bird's perspective, which is what you want to do now. The project was dealing with various road signalization, traffic signs and road markings. Useful references would be:

This should give you all the material you need to get what you want to do. Good luck, it took us a while.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy