Robotics StackExchange | Archived questions

Object detection - AR.Drone

Hello everyone,

In Gazebo, I have an AR.Drone 2.0 and a map, and I can see with both cameras, and print them using OpenCV with the code provided here:

http://wiki.ros.org/cv_bridge/Tutorials/UsingCvBridgeToConvertBetweenROSImagesAndOpenCVImages

In that code, a red circle is added to the original image. Now I want the camera to detect a predefined tag, for example:

image description

I understand that I have to modify the code above (the part where it draws the circle) with an algorithm that searches for that specific tag in the image and prints its result, but I don't know how to write that. One algorithm that I found is the TLD Predator (Tracking, Learning and Detection).

Thanks :D

Asked by Dylan on 2018-07-19 15:56:17 UTC

Comments

What is your question?

Asked by jayess on 2018-07-19 18:55:46 UTC

The question is: how can I modify that code to do what I said previously?

Asked by Dylan on 2018-07-19 20:35:43 UTC

This may not be, but it sounds like a homework problem.

Asked by jayess on 2018-07-19 20:45:39 UTC

What have you tried so far?

Asked by jayess on 2018-07-19 20:48:16 UTC

Nope. I’m working on a project (not for money!) and this is just part of it, but I’ve been reading about this the last 2 days and couldn’t find out exactly how to write it. I know how to read code in C++ but I have beginner level writing it. And writing a ROS code is even more difficult

Asked by Dylan on 2018-07-19 20:57:13 UTC

I read about some algorithms (for example, TLD Predator) and also saw some thesis (for example one named visual control for an ardrone, or something like that). I need to have a code that detect patterns so I can modify it and do what I need. This is just a starting point

Asked by Dylan on 2018-07-19 21:00:01 UTC

This is rather off-topic for this site. I'd say that you should ask on OpenCV Answers instead. Then, when you're ready ask about any issues that you're having with ROS on this site.

Asked by jayess on 2018-07-19 21:24:06 UTC

But I think that OpenCV people would say the same. The ROS part is more difficult, I can do (or investigate and study) the corresponding algorithm

Asked by Dylan on 2018-07-19 22:00:40 UTC

Well the ROS part of your problem is already solved with the example code you've provided. So now it's just image processing problem therefore nothing to do with ROS itself. If you want to detect chessboard you can have a look at camera_calibration package.

Asked by l4ncelot on 2018-07-20 02:51:07 UTC

On a side note, I noticed you are deleting your questions after some time. You have asked 9 questions of which 7 are deleted. This may discourage people to answer your questions as well as the fact that questions should be of help to other people that are experiencing similar problems.

Asked by pavel92 on 2018-07-20 05:50:47 UTC

Answers