True North Reference

[autresphere]

Hi Robby,

Do you have any idea on how the panorama can be truly referenced to the north ?

I've tried with [motionManager startDeviceMotionUpdatesUsingReferenceFrame:CMAttitudeReferenceFrameXTrueNorthZVertical] as well as CLLocationManager authorization and update. But each time the app is launch, and if you change the device azimuth, the panorama seem to have a different orientation.

In other words, let's say the sun in the image is exactly located on the east (meaning the sun is at the center of the screen when you look east with the device). Depending on how the device is oriented when the app is launched, the sun won't appear exactly at east. I thought CMDeviceMotion would take care of everything but I suspect an additional correction must be taken care of.

[mayakraft]

yep, I understand. This will be very useful. I'm looking into it now..

[autresphere]

I've asked Apple Technical support about this weird behavior, they answered this was a bug but unfortunately did not give any workaround.
Do you have any idea on how we could solve this?

[mayakraft]

@autresphere I develop on an iPod Touch, I had to dig up my old iPhone 4. I substituted this one line:

[motionManager startDeviceMotionUpdates]

with this:

[motionManager startDeviceMotionUpdatesUsingReferenceFrame:CMAttitudeReferenceFrameXMagneticNorthZVertical];

and I get a behavior which looks like it's working. It boots into default position (facing center of image), ignoring magnometer data, then after 0.5 - 2 seconds immediately rotates through the zenith to a compass-oriented azimuthal alignment.

I tested by booting the app with phone facing different directions, and after the first second, the landmarks were compass aligned in the same way.

Am I describing your issue correctly? My iPhone 4 is running iOS 7.1.1, not iOS 8. Unfortunately I can't do any kind of comparison, this is my only iOS device with a magnetometer on which to test.

[mayakraft]

If I got this correctly, and this is an iOS 8 bug, perhaps there's some data you can ask me to log, and we can compare logs?

[autresphere]

I'm actually not sure about this anymore. Sometimes, it seems to be always aligned in the same way (as you're saying), and sometimes it seems to be a little bit misaligned by several degrees, but the number of degrees seems far less important than what I could see. In the meantime, I've updated to iOS 8.1.3, maybe it did fix some inaccuracy.

[mayakraft]

hi @autresphere , I've been doing some work on this, I've run into another issue. Let's say the device is facing north when you boot, and you run 2 different builds - one with startDeviceMotionUpdatesUsingReferenceFrame:CMAttitudeReferenceFrameXMagneticNorthZVertical and one with just startDeviceMotionUpdates in both cases no further transformations are applied.

these result in 2 different orientation matrices, both have Z as the up vector orthogonal to Earth's ground (which I'm considering making into the new standard), but in the case of the reference frame, you have the forward vector XYZ (1,0,0), and with no reference frame XYZ (0,1,0). There becomes a 90deg discrepancy between where the center of the image is in both cases.

though in the case of a reference frame, during the split second after orientation begins but before the magnetometer boots up, they are both forward Y facing (0,1,0)

this would be easy to reconcile if there was a flag that flipped when the deviceMotion began incorporating magnetometer too, but none exist from what I've found (let me know if you find one), which leads me faced with rolling out my own code to apply magnetometer to deviceMotion.attitude, which maybe isn't too difficult, just not as elegant of a solution as I hoped.

[autresphere]

Hi @robbykraft, it seems you can check whether the magnetometer is active through CMDeviceMotion. magneticField.accuracy. A CMMagneticFieldCalibrationAccuracyUncalibrated value should tell you magnetometer is not yet incorporated. What do you think?

[sassiwalid]

i think that [Motion Manger startDeviceMotionUpdatesUsingReferenceFrame:CMAttitudeReferenceFrameXMagneticNorthZVertical] can resolve the pb of the orientation of the device with the orientation of the 360° image ???