使用 opencv/c++ 将像素转换为 mm --> 两点之间的距离

Question

I have two points p1(x1, y1) and p2(x2, y2). Both are centers of two circles in a 2D plan. The two circles are detected automatically in a picture taken by the user with his smartphone.

For some reasons I want to measure the distance (in meters) between p1 and p2.

I absolutely have no clue of how I can do it. I did have a look at calib3D module but I don't think I found anything useful.

The distance between the camera and the circles may vary and is nerver constant, I think this a parameter we could pay attention to? For the calculation of the distance of the two centers.

Is there a particular way to do such a thing? If you have any method that could help me, I'd be grateful.

Thanks, have a nice day.

Hugo

Answer 1

you need to use an Augmented Reality toolkit. nothing less will solve the problem, unless your problem has constraints that make it simpler than stated.

such toolkits are available for Apple and Android mobile platforms.

it makes no sense for you to try to implement the core of Augmented Reality yourself. some understanding of the theory does help in using a toolkit.

Answer 2

Ok so,

Here are the steps to A solution:

I) First make an orthonormal plan.

I needed two build up an orthonormal plan in which I will calculate the distance between the 2 points.

I need to know the distance in centimeters, I have a picture in pixels so I need to know how many centimeters is a pixel.

In my case, the picture is 2498 x 1570 pixels. When I measure with a ruler on the sheet taking the good corners:

2498 px = 18.2 centimeters and 1570 px = 11.4 centimeters.

With these datas we know that 1 px =~ 0.0073 centimeters. (18.2 / 2498 =~ 0.0073). (11.4 / 1570 =~ 0.0073)

II) Distance formula in an orthonormal plan

double Operation::distanceBetween2Points(Point A, Point B){

    double coeff = 0.0073;
    double diffx = (B.x - A.x) * coeff;
    double diffy = (B.y - A.y) * coeff;

    return sqrt(pow(diffx, 2) + pow(diffy,2));
}

I've found a distance equal to 0.61005 centimeters which matches the reality (= 0.6 cms).

I ran several tests, it works every times, matching the reality.