Visual MIMO

An Android implementation of visible light communication. See report here.

Setup - Android

• Import project in Android Studio
• Add: ndk.dir=[path/to/ndk] to local.properties
• Download folder OpenCV-2.4.10-android-sdk and place in root of project

Setup - MATLAB

• Run sample.m to generate vmimo.avi. It's short so run it looping: mplayer -loop 0 vmimo.avi.

NDK Code

C++ code is in app/src/main/jni/. The CV logic is in image_processing.cpp, with the NDK abstracted out. native.cpp connects image_processing.cpp to the Android code, and simulator.cpp allows developing CV algorithms without any Android by simulating the flow of images and corners from the Android code.

To run on a computer, cd to the jni directory and run make, then ./simulator.

Test Bench

The test bench is in bench (currently in the analytics branch). The project uses Maven, so you must right click on it > Configure > Make Maven Project (Eclipse, assuming Maven project installed). It utilizes VLCj to do video outputs, and directly references the server-side image generation.

Current hacks

• At the time of writing, Android Studio / gradle support for NDK is limited to projects that don't include libraries. The gradle config thus disables the NDK compile and manually runs ndk-compile when building the project.
• The Android project was based on an old Vuforia sample, so large parts of it are super ugly. If you want to add a UI element or feature, look at app/src/main/java/com/android/visualmimo/MainActivity.java.

TODO

• Play with new new embedding methods
• Clean up MATLAB code a lot
• Figure out how to draw on Android screen to display information
• Chromecast pairing demo