Why the 3d AR model (astronaut) is shaking so much?

[l2dev]

Why the 3d AR model (astronaut) is shaking so much?

[kannabi]

@eurbon @plattysoft had u found any solution?

[plattysoft]

The NASA logo does not have a lot of features, it is not a great choice of a marker. The model is quite large and tall, the shaking is way more noticeable with tall models, a small difference in angle on the marker (even if caused by a miscalculation) gets more noticeable when applied to a larger distance. All in all, this is inherent to the technology of ARToolKit, the marker and the model. All together amplify the problems. Yes, it was a very bad choice, but it was meant to be a quite simple example.

[kannabi]

It should be a very noticable miscalculating mistake because shaking is really big. Thank you for your answer anyway. It is useful)

[plattysoft]

It is not a miscalculation mistake as "the math is wrong"; it is due to jitter and image resolution quality + lack of features.

The algorithm can not detect the marker with enough accuracy if the marker has few features and the image resolution is not big enough. Some jitter is to be expected. In fact if you try to smooth it out using some easing algorithm you'll get bad accuracy and less responsiveness.

Same goes for having larger resolution images. They took longer to process, so the framerate and responsiveness gets impacted as well.

At some point you need to make a compromise of what is important for you and sacrifice some area. For the example, fps and responsiveness were considered more important than image stability.

And again, jitter is way more noticeable because the model is very large, it is amplified. A variation of 1 degree in otientation on 1 mm turns into 0.9998 or 0.0002 mm. If the model is 100 mm, the absolute error at the large end is 0.02. For the same error, the absolute value is much more noticeable.