Introduction

This page explains the complete build procedure for the uploaded examples.

Details

This work use OpenEmbedded based development environment. Following is the detail procedure to build this work.

Step 1

Setting up OpenEmbedded

First, download the OpenEmbedded setup scripts and save it.

# Click on this link http://gitorious.org/angstrom/angstrom-setup-scripts/commits/master

# Click on Download master as tar.gz. This will download and save file "angstrom-angstrom-setup-scripts-master-tar.gz".

# Double click on the file angstrom-angstrom-setup-scripts-master-tar.gz that you just downloaded.

# Then extract it to the folder of your choice. This folder will be called BASE_DIR.

Goto the folder BASE_DIR and run the provided scripts as mentioned below.

# cd BASE_DIR/angstrom-angstrom-setup-scripts

NOTE: This directory will be call OE_BASE. ie, OE_BASE = BASE_DIR/angstrom-angstrom-setup-scripts

# ./oebb.sh config beagleboard

# . ~/.oe/environment

The script to set-up OpenEmbedded environment uses 'git'. If you don't have git installed, you can install it using following command "sudo apt-get install git

# ./oebb.sh update

Now, you have a complete OpenEmbedded based development environment. If you want to build the kernel image and other necessary packages, you can simply type the following on the command prompt; "bitbake base-image" or "bitbake console-image" or "bitbake x11-image". For further details look into this link http://wiki.openembedded.org/index.php/Getting_Started.

Step 2

Building Codec-Engine

This project uses codec-engine 2.25.01.06 and it's dependencies, like DSPLink 1.64. Before you start bitbaking codec-engine, you need to download Code Generation Tools, ie, C6000 Code Generation Tools v6.1.9 for Linux from this link https://www-a.ti.com/downloads/sds_support/TICodegenerationTools/download.htm and save it to the directory OE_BASE/source/download. Then execute the following command from OE_BASE to build Codec-Engine and all of its dependencies.

# bitbake ti-codec-engine

The codec-engine and all other required tools will be build and saved at OE_BASE/build/tmp-angstrom_2008_1/sysroots/beagleboard-angstrom-linux-gnueabi/usr/share/ti. This process will also build the necessary kernel modules needed to work with on-chip DSP found on OMAP3530.

Step 3

Building OpenCV 2.1 With continious memory allocation support

It is assumed that you are at OE_BASE. Execute the svn check-out command to download the source of this project using following command.

# svn checkout http://opencv-dsp-acceleration.googlecode.com/svn/trunk/ opencv-dsp-acceleration

Now, you can see opencv-dsp-acceleration folder if you execute "ls" command.The patch is provided at opencv-dsp-acceleration/dsp_opencv/patch.

Since, we need to change the source code, it is preferred that you configure the OpenCV 2.1 and retain its source and configuration files. Following command can be used to do so.(It is assumed that you are at OE_BASE.)

# bitbake -f -c configure opencv

The above command will unpack and configure OpenCV 2.1 at OE_BASE/build/tmp-angstrom_2008_1/work/armv7a-angstrom-linux-gnueabi/opencv-2.1.0+svnr3217-r1. Lets call this OPENCV_BASE.

Note:- Name of folder opencv-2.1.0+svnr3217-r1 can vary depending on svn revision. In this case the numbers after +svnr can be different.

Now change directory to OPENCV_BASE.ie, OE_BASE/build/tmp-angstrom_2008_1/work/armv7a-angstrom-linux-gnueabi/opencv-2.1.0+svnr3217-r1 .

# cd OE_BASE/build/tmp-angstrom_2008_1/work/armv7a-angstrom-linux-gnueabi/opencv-2.1.0+svnr3217-r1

Execute following commands to apply the patch

# patch -p0 < OE_BASE/opencv-dsp-acceleration/dsp_opencv/patch/opencvdsp.diff

Execute following command to add include directory of CMEM library to the CXX_FLAGS in the files OPENCV_BASE/modules/core/CMakeFiles/opencv_core.dir/flags.make and OPENCV_BASE/modules/core/CMakeFiles/opencv_core.dir/flags.make.

# sed -i '/^CXX_FLAGS/s:$: -I**<OE_BASE>**/build/tmp-angstrom_2008_1/sysroots/beagleboard-angstrom-linux-gnueabi/usr/share/ti/ti-linuxutils-tree/packages/ti/sdo/linuxutils/cmem/include:g' opencv/modules/core/CMakeFiles/opencv_core.dir/flags.make

Similarly,

# sed -i '/^/s:$: <OE_BASE>/build/tmp-angstrom_2008_1/sysroots/beagleboard-angstrom-linux-gnueabi/usr/share/ti/ti-linuxutils-tree/packages/ti/sdo/linuxutils/cmem/lib/cmem.a470MV :g' opencv/modules/core/CMakeFiles/opencv_core.dir/link.txt

NOTE: Don't forget the space after '$:' in the commands above. Also expand OE_BASE to real path.

Now, change the directory to OE_BASE and then bitbake opencv

# cd OE_BASE

# bitbake opencv

Note: The above patch effects only libopencv_core.so, libopencv_core.so.2.1 and libopencv_core.so.2.1.1. If you already have a working copy of opencv, you can copy only these library files related to libopencv_core to your beagleboard at /usr/lib. These libraries can be found at OPENCV_BASE/lib.

If you don't hava a working copy of OpenCV 2.1, install all OpenCV packages generated by bitbake on Beagleboard.

Step 4

Building This Project

To build this project, go to the folder you just checkedout from SVN. ie,

#cd OE_BASE/opencv-dsp-acceleration/dsp_opencv

Edit file Rules.make to point to the correct path for each tools. In most of the case editing varialble ROOTHPATH is sufficient, if you have followed the method described in step 2 to install Codec-engine and tools. Also verify the path of cross-compiler in the variable CROSS_COMPILE to point to correct path. It is advisable that you edit EXEC_DIR variable to point to the folder where you want all of the executables and test files to be saved after the project is built. If you are using NFS to share folder with Beagleboard, the EXEC_DIR variable can also be the path to this folder.

Note: If you have VLIB and CXCORE library, you need to uncomment the line in file Makefile at OE_BASE/opencv-dsp-acceleration/dsp_opencv/beagle/c6accel_1_00_00_04_async/dsp/alg/pjt. Which can be done as shown below.

Before ```

uncomment the lines below if you are using VLIB. The library file must be stored at [c6accel_Base_Dir]/dsp/libs

VLIB = -l"../../libs/vlib.l64p"

uncomment the lines below if you are using CXCORE. The library file must be stored at [c6accel_Base_Dir]/dsp/libs

CXCORE = -l"../../libs/cxcore.lib"

```

After ```

uncomment the lines below if you are using VLIB. The library file must be stored at [c6accel_Base_Dir]/dsp/libs

VLIB = -l"../../libs/vlib.l64p"

uncomment the lines below if you are using CXCORE. The library file must be stored at [c6accel_Base_Dir]/dsp/libs

CXCORE = -l"../../libs/cxcore.lib" ```

Now,

Execute following command to build the project

#make all

This command will build the DSP executable c6accel_omap3530.x64P, Arm-side shared library lib/libopencvdsp.so and Arm executable remote_ti_platforms_evm3530_opencv.xv5T. The two executables and lib folder will be located at EXEC_DIR. Now,copy c6accel_omap3530.x64P, remote_ti_platforms_evm3530_opencv.xv5T to the Beagleboard at your work directory. Lets call this directory WORK_DIR. Copy libopencvdsp.so to the Beagleboard at /usr/lib. Also, copy the movie clip tree.avi found on the EXEC_DIR folder to the WORK_DIR.

Step 5

Restart Beagleboard and Set-up Memory

Set Memory for Linux

Before booting up the beagleboard, set aside memory for linux from uboot prompt as shown below.

OMAP3 beagleboard.org # setenv bootargs console=${console} vram=${vram} mapfb.mode=dvi:${dvimode} omapfb.debug=y omapdss.def_disp=${defaultdisplay} root=${mmcroot} rootfstype=${mmcrootfstype} mem=80M@0x80000000 128M@0x88000000

Copy Kernel Modules for DSPLink

Copy the kernel modules to the WORK_DIR on Beagleboard. These modules can be found at OE_BASE/build/tmp-angstrom_2008_1/sysroots/beagleboard-angstrom-linux-gnueabi/lib/modules/2.6.32/kernel/drivers/dsp.

Load all the kernel modules, dsplinkk.ko, lpm_omap3530.ko and cmemk.ko.

You can load these modules using the script loadmodules.sh provided in the OE_BASE/opencv-dsp-acceleration/dsp_opencv/scripts. To unload the modules, unloadmodules.sh is provided in the same folder. Copy these files to your WORK_DIR in Beagleboard and type following

$ ./loadmoudles.sh

You can check whether the modules are loaded or not by using lsmod command.

Also, it is suggested that you load the uvc module since the application for demonstration of this project uses webcam.

Step 6

Execute

Type the following command at the command prompt on Beagleboard to run the application. This will show the usage of the command. You can then follow the usage after that.

$cd WORK_DIR

$./remote_ti_platforms_evm3530_opencv.xv5T

Note

1) To test the example you may need to copy the test data file tree.avi as mentioned above or need to have a webcam. I have tested the app using Logitech webcam Pro 9000 to demonstrate Sobel algorithm.

2) The example demonstrates 2-D 63x63 forward DFT. The output data is displayed on the screen.

3) The example also demonstrates 3x3 Sobel algorithm for frames captured using webcam. Due to low visibility of signed-image when using cvShowImage(), the image is converted back to unsigned and then displayed.

5) The example also demonstrates 3x3 Sobel algorithm for input movie clip tree.avi.

6) Integral algorithm is also demonstrated. You need to install VLIB for DSP based cvIntegral()