Search Unity

Question ARKit GeoAnchors functionality with ARKit image anchors

Discussion in 'AR' started by ReallyRad, May 13, 2021.

  1. ReallyRad

    ReallyRad

    Joined:
    Dec 19, 2018
    Posts:
    3
    Hi,

    I'm working on a simple experience where some AR content must be placed relative to an actual city landmark. This is basically the GeoAnchors functionality from ARKit that I need, unfortunately, it is not yet available in Europe. I was wondering if it could be possible to achieve something similar with Image anchors. The idea would be to treat the building facade or part of the facade as an Image Anchor.

    I want to avoid an image marker on the floor in the buildings surroundings because in my experience this is not so precise and tends to drift once you lift the phone and stop detecting the floor. Also any angular drift will be quite bad when multiplied by the distance to the facade.

    So my question is : can I use image recognition on a building facade? Are there limits to image size recognition?

    thanks
     
  2. corto_maltese

    corto_maltese

    Joined:
    Jun 1, 2021
    Posts:
    1
    Hey ReallyRad,

    maybe this isn't topical anymore, … but have you looked into ARTrackedObjectManager and reference objects? Not sure if that works for ARCore/Android. But you sound like you might have an iOS-only solution in mind. Apple provides documentation and useful companion app code for this.

    Of course, a building's facade or parts of it can be fairly 2D and have enough detail to work as an image anchor. I'm doing a scavenger hunt in the park project with a wild mix of (parts of) buildings, sculptures, a 2D sign with the park map. Surprisingly (or maybe not), a reliant type of 3D anchor turned out to be tree trunks.

    I built a 2-layer positioning-anchor system, with the basic layer being device sensors (GPS and compass), and the occasional 3D or 2D scene anchor as the optional higher-quality layer.

    As long as ARGeoAnchors are not available, I think it's necessary to check whatever existing feature (3D, 2D, GPS/compass) works best in each particular spot.
     
  3. mikeyrafier98

    mikeyrafier98

    Joined:
    May 19, 2022
    Posts:
    37
    Actually, I'm quite interest in your story.
    You mentioned about use 2-layer-tracking: 1) from GPS + device sensors, 2) from 2D landscape/3D object tracking
    What did you do in order to sync both data?