Wikitude
How can we help you today?
problem with getTargetPositionInCameraFrame
Wikitude Technical Support
Hi Laurent,
are you rendering the camera yourself or do you let the Wikitude SDK render it?
How do you plan to set the focus area in your camera implementation?
Best Regards,
Alex
Hi Alex,
I let the SDK render the camera frame itself (pretty close to simpleInputPlugin sample)
I set the focus area in my implementation of wikitudeCamera.java, using camera1 api. In my plugin I call getTargetPositionIncameraFrame (in update callback), and give this coordinates to my camera class.
I tried to draw the rect returned by this api on the screen but as the resolutions don't match, I'm forced to convert the rect into a 1280x720 area, but the rect dosn't follow the target very well when its seen with angles or when moving closer to the target.
Please see attached video. (the purple rect is the targetPositionInCameraFrame)
Regards,
Laurent
Laurent
Wikitude Technical Support
Hi Laurent,
what you need to consider when calculating the position on the screen is that the parts of the camera frame are not visible. This is because we scale the camera frame to fit the view. In the case of a 4:3 camera frame being displayed on a 16:9 view, the complete width of the frame is visible but part of the height is not visible.
Best Regards,
Alex
That's what I was supposing.. In the video we can see that the target vertical position and height are correct, but the target width and horizontal position are not good.
Please find also attached the code we use to convert the wikitude rect into a screen rect. (maybe we didn't anderstood something..?)
Regards,
Laurent
Wikitude Technical Support
Hi Laurent,
please take a look at this pseudo-code which could work for your case.
targetRect // assume 200x100 rect in the center x = 320 y = 240 w = 200 h = 100
displayRatio = displayWidth / displayHeight // assume 1920x1080 -> 1.77777 device in orientation landscape
frameRatio = frameWidth / frameHeight // assume 640x480 -> 1.33333
if displayRatio > frameRatio
scaledDisplayHeight = displayWidth / frameRatio // -> 1920 / 1.3333 = 1440
scale = displayWidth / frameWidth // -> 1920 / 640 = 3
diff = (scalescaledDisplayHeight - displayHeight) / 2 // -> (1440-1080) / 2 = 180 this is the height of the part that was cut off on both sides
x = targetRect.origin.x * scale // -> 320 * 3 = 960
y = targetRect.origin.y * scale - diff // -> 240 * 3 - 180 = 540
w = targetRect.size.width * scale // -> 200 * 3 = 600
h = targetRect.size.height * scale // -> 100 * 3 = 300
scaledTargetRect({x,y},{w,h})
It scales the rectangle you get by getTargetPositionIncameraFrame up to the display size.
Best Regards,
Alex
But I found the same issue on iOs, using SDK 7. Did you report this fix also on iOs SDK ?
Regards,
Laurent
Wikitude Technical Support
Hi Laurent,
which issue did you find on iOS? The issue with the 720p camera or the one fixed with the pseudo-code?
The 720p issue was not platform specific and should be fixed.
For the other issue you will still have to apply the same code as you did in android.
Best Regards,
Alex
The issue I found is about the 720p: the coordinates returned by getTargetPosition seems to be in a 640x480 reference whereas the camera frame size is 720p.
Regards
Laurent
Wikitude Technical Support
Hi Laurent,
please try to use the pre-release version of the Wikitude SDK 7.1. It can be downloaded in the download section under the beta category.
Best Regards,
Alex
Laurent Latour
Hi,
I'm using the SDK 6.1, with an inputPlugin on android.
We need to get the exact position of the target in the camera frame, in order to autofocus the camera in a special area of the target.
I know there's a bug in this release about using it with a 720p resolution, (see post https://support.wikitude.com/support/discussions/topics/5000083211) but I still have a problem with this api:
- When the SDK is configured to use 640x480 camera frames (so 4:3 format), and the camera preview size is 640x480 (like in your samples), the viewfinder is displayed fullscreen (so in a 16:9 format). I can't figure out how to convert the values returned by getTargetPositionInCameraFrame() from a 4:3 frame to a 16:9 view..
I can't find any documentation about this on your website.. could you explain how it is supposed to work ?
Thanks,
Laurent