Leverages OpenCV through Wi-Fi to accomplish AR.Drone 2.0 swarming behavior.
The application is currently built with C++ using OpenCV 2.4.10 through any OS or IDE that successfully compiles it without interfering with Git.
With successful compilation of the application, one can add an "object.jpg" file to the executable's directory to match a specific object against the default webcam stream through homography. Success of the the match is denoted by rectangular encapsulation of the object in thick green lines.
The eventuality is to refactor the webcam upstream to the drone's video port upstream after feature recognition is worked out more extensively. Then, every so often, flight commands can be sent to the follower drone based on the recognized features from the leader drone by judging the distance from the absolute center of the following drone's vision. Basically, the follower drone should always attempt to center its vision on the leader drone.
Depth will be accounted for by calculating the size of combined recognized features, always attempting to scale to the known features appropriately.
AR.CV adapts this tutorial from the official OpenCV documentation.
Here's a tutorial for building OpenCV on OS X. The guide works with Yosemite. Note that CMake can be installed via Homebrew.
This guide is also confirmed to work with XCode 6.1.1.
Here's a tutorial on installing OpenCV for Windows by use of an official binary file.
The guide works with Windows 8.1.
This guide is also confirmed to work with Visual Studio 2013. Just use the "VS12" linkage built into OpenCV -- it works with Visual Studio 2013. Be sure to use the OpenCV 2.4.10 library files as per the system requirements.
Follow this tutorial to install. Tested and works on Debian Wheezy.