Need help finding convertion ratio px/cm

2ant

Hi, i have an OAK-D pro poe with the following specs:

-1280*720px

-HFOV 69°

-VFOV 55°

I'm detecting a box that is 187pixels by 238 pixels.

The real (measured) dimensions are 32.5cm by 43.5cm.

That give two ratio:

43.5/238=0.1828 cm per pixel

32.5/187=0.1738 cm per pixel

I know that the box is 1.55m from the camera.

I'm trying to find a way to measure the object throught code.

I tried to do this :

55°/720=0.0764 ° per pixel

69/1280=0.0539 ° per pixel

i don't know what to do next.

jakaskerl

Hi 2ant
The spatial calculation example should guide you to get the dimensions of an object; namely, the calc.py script, which has the x, y, z calculator for each point on the depth frame.

If you have any more questions, feel free to ask.

Thanks,
Jaka

2ant

My goal is to measure the area of the object which won't always be a nice rectangle.

I do not know how to find the answer with what is in that file (calc.py).

I also did this:

187*238=44506 (area in pixels)

44506*(0.1828*44506)*(0.1738/44506)=1413.98 which is VERY close to the real area of the object (1413.75)

If someone knew how to find those conversion factor it would be awesome.

I don't have enough math knowledge to find them.

Whatever Chatgpt or i try is nowhere close to the real results.

2ant

I also tried using this : https://fov.luxonis.com but it doesn't give me correct values.

2ant

The 187 and 238 are the dimensions of the red bounding rectangle.

jakaskerl

Hi 2ant

Calculate x and y (in pixels) relative to the center of the image.

midW = int(depthFrame.shape[1] / 2) # middle of the depth img width
midH = int(depthFrame.shape[0] / 2) # middle of the depth img height
x_offset= x - midW # x in pixels
y_offset = y - midH # y in pixels

This will give you the angle (in radians) between the point and the center of the image.

def calc_angle(frame, offset, HFOV):
        return math.atan(math.tan(HFOV / 2.0) * offset / (frame.shape[1] / 2.0))

Use the function to calculate angles.

# Required information for calculating spatial coordinates on the host
HFOV = np.deg2rad(self.calibData.getFov(dai.CameraBoardSocket(depthData.getInstanceNum())))

angle_x = self._calc_angle(depthFrame, x_offset, HFOV) # angle in x direction
angle_y = self._calc_angle(depthFrame, y_offset, HFOV) # angle in y direction

Then acquire the x, y and z component of a point on the image.

z = depth,
x = depth * math.tan(angle_x),
y = -depth * math.tan(angle_y)

Hope this helps,
Jaka