4

Can anyone tell me what exactly or how exactly does an image get recognised in the Qualcomm QCAR https://developer.qualcomm.com/develop/mobile-technologies/augmented-reality. I would like to know which image detection algorithm is used.

Misty
  • 61
  • 2
  • 4

2 Answers2

6

QCAR is based on Studierstube ES, which is publically not available. However, you might be interested in publications of the handheld augmented reality project as some of those are about Studierstube ES, the predecessor. The algorithms for sure have been improved in the meantime, but you can find out from what they have evolved.

I didn't read the the papers, but these could be interested to you:

http://www.icg.tugraz.at/publications/pdf/WAGNER_ISMAR08_AMT.pdf

http://www.icg.tugraz.at/publications/pdf/WAGNER_ISMAR08_NFT.pdf

http://www.icg.tugraz.at/Members/daniel/MultipleTargetTrackingPDF

http://www.icg.tugraz.at/Members/daniel/Publications/TVCG2010_RealTimeDetectionAndTrackingOnMobilePhones.pdf

Pedro
  • 4,100
  • 10
  • 58
  • 96
  • Thanks pedro.. I hope I find lot of useful info from the links that you have mentioned. Will surely get back on to u on this. – Misty Nov 15 '11 at 07:28
1

Unfortunately there seems to be little to no documentation about the applied algorithms in QCAR on the Qualcomm website, other than "The QCAR SDK uses sophisticated algorithms to detect and track the features that are naturally found in the image itself"

However, there is some explanation about the different types of supported trackables you might want to take a look at. I'm not much into image recognition, but you may be able to get some pointers about the underlying algorithms from that.

MH.
  • 45,303
  • 10
  • 103
  • 116
  • Thanks MH. By the way I did check out the different types of trackables as you have mentioned. But my main concern is how do they do Image recognition. I have already created the image trackable, but if I place 2 similar images ( that differ in few features) even those are getting recognized as the same. So if I come to know of the Image detection algorithm, I can precisely design my application. Please suggest – Misty Nov 04 '11 at 12:38
  • It appears Qualcomm launched [FastCV](https://developer.qualcomm.com/blog/introducing-fastcv-computer-vision-technology-tuned-mobile) less than two weeks ago - the computer vision library that also sits at the heart of QCAR. I haven't gone through the whole [documentation](https://developer.qualcomm.com/docs/fastcv/api/index.html), but that may be worth a shot for more details regarding the image recognition. I would also suggest you submit your question on the [FastCV developer forum](https://developer.qualcomm.com/forums/qdevnet-forums/computer-vision-fastcv). – MH. Nov 05 '11 at 19:25
  • Thanks MH. I shall go through it and will surely update if I get any details. – Misty Nov 10 '11 at 04:13