0

I am interested in developing an Android Application that employs the Android Devices Camera to detect moving "Targets".

The three types of targets I need to detect and distinguish between are pedestrians, runners (joggers) and cyclists.

The augmented realities SDK's I have looked at only seem to offer face recognition which doesn't sound like they can detect entire people.

Have i misunderstood what Augmented Realities SDK can provide?

LOG_TAG
  • 19,894
  • 12
  • 72
  • 105
Hector
  • 4,016
  • 21
  • 112
  • 211
  • thank you SO once again for a down & close vote without explanation. how does that behaviour help any one? what value is there in un explained close and down votes? i have a genuine need for assistance and guidance from an expert community. In return that expert community can spare their time to only downvote and/or close questions. – Hector May 23 '15 at 08:40

1 Answers1

1

There is a big list of AR SDKs (also for Android platform):

Augmented reality SDKs

However, to be honest I strongly doubt that you will find any (doesn't matter free or payed) SDK for your task. It is to specific so you should probably write it by yourself using OpenCV.

OpenCV will allow you to detect objects (more or less) and then you will need to write some algorithm for classification. I would recommend classification based on object speed.

Then, when you have your object classified you can add any AR SDK to add something to your picture.

sebap123
  • 2,541
  • 6
  • 45
  • 81
  • as far as I know opencv is quite "heavy". Depending on the type of the AR application (e.g.: real time detection/classification) opencv might not be the best solution unless the processing is performed on the server-side of course.. – Traian May 25 '15 at 00:07
  • @sebap123, if only i could get the openCV motion detection code and/or app to work :( – Hector May 26 '15 at 12:16
  • @unichiduci, what other options do i have? from what i can see i either use openCV or do it myself by comparing two images captured with a time delay, reduce them then compare – Hector May 26 '15 at 12:18
  • yes Hector, depending on the complexity of your algorithm, sometimes it's better to use your own code which does exactly what it is supposed to do. Using existent libraries which are not created for a specified use-case might considerably slow down the overall processing time which is not what you want in an real-time AR application. Of course you can try it out before and if the results are acceptable you can use OpenCV or any other existent library. If you're not happy with the result then you can implement your own solution. It depends very much on your time availability :) – Traian May 29 '15 at 14:30