Log in or register to post comments

How Should One Choose An Appropriate Target Type?

October 11, 2012 - 7:38am #1

Hi,

I'm new to AR and the QCAR API, and I'm hoping someone might be able to cut through a bit of the noise, and help me decide which sorts of Trackables I should be using.

I'm part of a small research project that aims to augment buzzer boxes (the kind you might find in front of an apartment building) so that delivery workers can see an AR view of resident's preferences by simply scanning the box with a smartphone.  (Here's an example of the kinds of buzzer boxes we have in Copenhagen. https://docs.google.com/open?id=0B0quuhB29cakdEtUU3c0cmNNeUE). 

Ideally, when the delivery worker views the box through the phone's camera, we'd like to be able to see an AR icon - such as a checkmark - hovering over each buzzer button. This icon represents whether a resident is willing to accept deliveries or not.

PROJECT CONSTRAINTS

- Support for Multiple Targets: Some buildings have 8 - 12 seperate residences, and I need to be able to track them all, since they all have independent preferences.

- Recognize Hundreds/Thousands of Unique Buzzer Boxes: Ideally, it would be nice to be able to have a database which contains information for all of Copenhagen's apartment buildings.

- Dynamic / Realtime Recognition: We can not realistically know, in advance, exactly what every buzzer box will look like. Instead, we need a system that can recognize and decode an arbitrary image. (As can be done with QR Codes)

- Small Size: Many of the buzzer boxes have a large swath of blank steel where a large format image/glyph can be posted, but the buttons themselves (which I assume will also need to be tagged) can be quite small (i.e. A diameter of 16 - 20 millimeters, or 0.6 - 0.8 inches.)

MY UNDERSTANDING SO FAR...

- Based on my reading so far, it is my understanding that the QCAR API is limited to 50 Image Targets. 

- Image Targets must be known (and analyzed) in advance.

- Multi Targets must be known in advance. For these reasons, it seems that Image and Multi Targets aren't the way to go.

- It seems that Frame Markers can be decoded dynamically, but they only support 512 unique codes. 

- In my experiments with Frame Markers, they don't work very well when they are very small. 

- All Trackables provide Pose information.

SOME SPECIFIC QUESTIONS...

- Is there some kind of Trackable image which is both dynamically recognizable AND capable of being small?

How Should One Choose An Appropriate Target Type?

June 24, 2013 - 9:26am #4

The cloud database only supports single Image Targets.

Feel free to add this to the wish list.

 

N

How Should One Choose An Appropriate Target Type?

June 24, 2013 - 6:24am #3

How it is possible to use multitargets using vuforia cloud .cloud database allows only image .There is no option to add the cube with different faces in the cloud database.

 

Please let me know how to make it possible.'

Thanks

 

How Should One Choose An Appropriate Target Type?

October 11, 2012 - 3:33pm #2

Our Cloud Recognition service is capable of recognizing very large sets of unique images, and providing trackable datasets for these - See: https://developer.qualcomm.com/mobile-development/mobile-technologies/augmented-reality

With Cloud Reco you can store and retrieve both ImageTargets and MultiTargets.

Have you considered identifying boxes by GPS location so that you can reduce the number of unique targets? (e.g. https://gimbal.com/ integrates Vuforia and location ).

Also you might consider using a target placed in a consistent position relative to the buttons to serve as an anchor for the augmentation, rather than attemping to use the button faces. Their size, and the fact that they'll experience corrosion and wear are going to present a challenge for any target or marker system. A good ImageTarget a/o MultiTarget configuration will allow your to offset the augmentation from the target a bit, which may enable you to design a frame for the button console so that the augmentation can be rendered over the buttons.

 

 

 

Log in or register to post comments