EastBay RC Drone Olympics Stuff. This is a thing we put together for the SF droneolympics.com. We'll tidy up after the contest.
ArctiNode -- Web-page based ar.drone control, video capture and navigation.
Team EastBay RC
Mark Harrison ([email protected]) Andreas Oesterer ([email protected]) http://eastbay-rc.blogspot.com
ArctiNode is a node.js application for controlling an ar.drone via a web browser.
Commands directly supported: Takeoff, Land, Freeze, Reset Emergency Mode, Start Video Stream
Dual HTML 5 virtual joystick controller interface. The controls are like traditional RC transmitters: left stick controls altitude and yaw (rotation), right stick controls motion (left/right/forwards/backwards).
Video Capture. Video frames from the ar.drone are captured to the host computer and displayed in the browser.
Facial Recognition. Frames captured are sent to OpenCV for facial recognition.
Light Recognition. Recognizes a light source such as a flashlight or laser pointer. [incomplete -- having problems extracting R/G/B channels from video stream]
PID controller. A general purpose feedback control loop for smoothly controlling ar.drone motion. Can be used for other applications that need a PID controller. A PID testbed is included.
Facial Following. The drone will turn to point to anything that it recognizes as a face. The facial recognizer calculates how far away the face is from being centered, and feeds that value as the error value to the PID controller, which then rotates the ar.drone. For safety's sake the ar.drone does not move towards the face. [incomplete -- we will try to hook this up during the contest.]
Drone Following. The ar.drone will hover over another unit and track it's movements using an LED beacon attached to the other unit [incomplete: depends on Light Recognition and downward facing video capture as noted above and below.]
Of course, a Unix command line interface with convenient vi-compatible mnemonics is also provided for those who prefer flying aircraft the old fashioned way.
Things we are looking for help with. Any assistance much appreciated!
-
Video stream consists of 3 8-bit channels packed into 4 bytes and cast into floating point. How can we extract these 3 values from the floating point value returned to javascript?
-
How can we call the OpenCV threshold functions?
-
Capturing video at too high a frame rate causes node.js errors.
-
Is there an existing ardrone call to stream video from the downward-facing camera? There's a low-level command documented in the Parrot reference.
NODEJS
- node-v0.8.15.tar.gz
- configure --prefix=/Users/mh/nodejs
- make; make install
OPENCV
- OpenCV-2.3.1a.tar.bz2, as per nodejs opencv docs
- edit CMakeLists.txt
- cmake -G "Unix Makefiles"; make; make install
- export DYLD_LIBRARY_PATH=/Users/mh/nodejs/lib
ARDRONE
- npm install git://github.com/felixge/node-ar-drone.git -g
NODEJS OPENCV
- npm install opencv -g
- ?? export PKG_CONFIG_PATH=/Users/mh/nodejs/lib/pkgconfig
ENV
- export DYLD_LIBRARY_PATH=/Users/mh/nodejs/lib
- export PATH=/Users/mh/nodejs/bin:$PATH
- ?? export PKG_CONFIG_PATH=/Users/mh/nodejs/lib/pkgconfig