What is a recommended way to find the corresponding points of the top-down image when calculating the Perspective Transform?

asked 2017-02-21 10:00:45 -0600

sjuneja
6 ●2

Is it done just visually?

Like here http://www.pyimagesearch.com/2014/08/...

image description

So if it was a more complicated image like that of a road, and I wanted a top-down view of part of a lane of the road. The only way for me to select corresponding points for the best possible top-down view would be by visual inspection?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2017-02-21 13:55:14 -0600

LBerger
9317 ●2 ●20 ●88 http://www.traimaocv.fr

Visually what do you mean ?

In your example four green points are supposed to be a rectangle. getPerspectiveTransform find a transformation between 4 green points and a rectangle. You have to choose a rectangle.

You have to know original shape in your image to find the transformation...

In road context try to find some known shape like traffic light ....