Kinect Installation on BeagleBoard-xM and PandaBoard running Ubuntu 12.04

Important Note

SyntroKinect and the Kinect infrastructure use a significant amount of processing power and, while it can be used on the BeagleBoard-xM and Pandaboard, performance can be quite poor.

Prerequisites

These instructions have been tested for Ubuntu 12.04 loaded from here. Select the appropriate Texas Instruments OMAP3 (BeagleBoard-xM) or OMAP4 (PandaBoard).

The Kinect driver and middleware need to be built from source and there are some packages required. Enter:

sudo apt-get update
sudo apt-get install gcc-multilib libusb-1.0.0-dev git-core build-essential
sudo apt-get install doxygen graphviz default-jdk freeglut3-dev libopencv-dev

Installing OpenNI

Download the latest unstable version of the OpenNI software from GitHub:

cd ~/
mkdir kinect
cd kinect
git clone git://github.com/OpenNI/OpenNI.git
cd OpenNI
git checkout unstable

Take a look at the README. This should say that the version is at least 1.5.4.0.

The software is set up for software floating point but the Ubuntu has been compiled for hardware floating point. So, the compiler flags need to be modified. To do this, enter:

cd Platform/Linux/Build

Edit Common/Platform.Arm. The original looks like this:

ifeq "$(CFG)" "Release"

    # Hardware specifying flags
    CFLAGS += -march=armv7-a -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp #-mcpu=cortex-a8

    # Optimization level, minus currently buggy optimizing methods (which break bit-exact)
    CFLAGS += -O3 -fno-tree-pre -fno-strict-aliasing

    # More optimization flags
    CFLAGS += -ftree-vectorize -ffast-math -funsafe-math-optimizations -fsingle-precision-constant

endif

Remove the “-mfloat-abi” so that it looks like this:

ifeq "$(CFG)" "Release"

    # Hardware specifying flags
    CFLAGS += -march=armv7-a -mtune=cortex-a8 -mfpu=neon #-mcpu=cortex-a8

    # Optimization level, minus currently buggy optimizing methods (which break bit-exact)
    CFLAGS += -O3 -fno-tree-pre -fno-strict-aliasing

    # More optimization flags
    CFLAGS += -ftree-vectorize -ffast-math -funsafe-math-optimizations -fsingle-precision-constant

endif

To build, enter the following:

cd ~/kinect/OpenNI/Platform/Linux/CreateRedist
./RedistMaker.Arm

It’s possible that this will result in an error because MAKE_ARGS includes “-j0″ which is illegal. If this occurs, edit Redist_OpenNI.py and find the line that looks like:

MAKE_ARGS += ' -j' + calc_jobs_number()

and change it to

MAKE_ARGS += ' -j1'

This will build the OpenNI binaries and create a folder called Redist. So enter the following to install:

cd ~/kinect/OpenNI/Platform/Linux/Redist/OpenNI-Bin-Dev-Linux-Arm-v1.5.4.0
sudo ./install.sh

The version number in the path may need to be changed to reflect whatever is the current version.

Installing the Driver

Download the source code like this:

cd ~/kinect
git clone git://github.com/avin2/SensorKinect.git

The software floating point option has to be turned off. Enter this:

cd ~/kinect/SensorKinect/Platform/Linux/Build/Common

Edit the Platform.Arm file to remove the “-mfloat-abi” option as for OpenNI.

To build, enter:

cd ~/kinect/SensorKinect/Platform/Linux/CreateRedist
./RedistMaker

This will build the driver and create a Redist folder. Enter this:

cd ~/kinect/SensorKinect/Platform/Linux/Redist/Sensor-Bin-Linux-Arm-v5.1.2.1

There’s an edit required to a config file or else it will not select the correct USB port. Edit Config/GlobalDefaultsKinect.ini. There’ll be a line that looks like:

;UsbInterface=2

Change this line to be:

UsbInterface=1

Note that the line has been uncommented and the interface has been changed from 2 to 1. Now the driver can be installed. Enter:

sudo ./install.sh

Testing

It’s worth trying to run a sample to see if this has worked. Enter:

cd ~/kinect/OpenNI/Platform/Linux/Redist/OpenNI-Bin-Dev-Linux-Arm-v1.5.4.0
cd Samples/Bin/Arm-Release
./Sample-NiSimpleRead

This should result in a series of values being displayed which vary if something is waved in front of the Kinect. In principle, it should also be possible to run:

java -jar org.OpenNI.Samples.SimpleViewer.jar

but this crashes at the moment. Non-GUI samples seem to work though.

It’s possible that the “gspca_kinect” problem will cause the sample not to see the Kinect. If the sample doesn’t run, enter:

sudo rmmod gspca_kinect

and try again. It looks like the driver install blacklists this module so an alternative would just be to restart the BeagleBoard or PandaBoard.

출처 : http://www.pansenti.com/wordpress/?page_id=1772