Category Archives: linux

The Case of OpenCV and the Missing SURF

I have been wanting to have a play with the OpenCV computer vision framework on Python for a while and finally got some time to some experimenting, I am looking to have the computer recognise LEGO parts, after much research and mucking about it seems I should be using cv2.SURF and/or cv2.SIFT for what I want to do. However on Fedora 19 these are not included in the distribution RPM as they are nonfree in that they are not open source. Attempts to use SIFT or SURF result in the following error:

$ python
Python 2.7.5 (default, Nov 12 2013, 16:18:42)
[GCC 4.8.2 20131017 (Red Hat 4.8.2-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
>>> print cv2.__version__
2.4.6.1
>>> i = cv2.SURF()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
AttributeError: 'module' object has no attribute 'SURF'
>>>
This is annoying as OpenCV will now need to be re-installed the old fashioned way, but there are instructions on the OpenCV site and I will be using them here with additional information I have discovered while following them.

Change to root and add the rpmfusion.org free and nonfree repositories as ffmpeg and libv41 are unavailble from Fedoras, followed by an update (it seems that Fedora 19 always has something to update):

& sudo su
# yum localinstall --nogpgcheck http://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm http://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm
# yum update

Remove the existing OpenCV packages:

# yum erase opencv*

There are a few packages to install in preparation to compiling the code, you may have some of all of these already but yum will skip those. Install the mandatory packages, these are for compiling the source and using GTK for the GUI:

# yum install cmake python-devel numpy gcc gcc-c++ gtk2-devel libdc1394-devel libv4l-devel ffmpeg-devel gstreamer-plugins-base-devel git wget

Install the optional dependencies, I have gone for the install everything approach, they may be useful later:

# yum install libpng-devel libjpeg-turbo-devel jasper-devel openexr-devel libtiff-devel libwebp-devel tbb-devel eigen3-devel python-sphinx texlive

Now, we are ready to get the latest source, here there are two ways to do this, install the latest development version using GIT, or download the latest stable version. My preference is to use the latest stable version.

With GIT, exit back to yourself from root, change to your home directory and download OpenCV:

# exit
$ cd ~
$ git clone https://github.com/Itseez/opencv.git
Cloning into 'opencv'...
$ cd opencv
$ mkdir build
$ cd build

Or using the latest build, at time of writing this is v2.4.7. Get this via the downloads page and save it to your home directory:

# exit
$ cd ~
$ tar -zxvf opencv-2.4.7.tar.gz
$ cd opencv-2.4.7
$ mkdir build
$ cd build

Now configure:

$ cmake -D CMAKE_BUILD_TYPE=RELEASE -D BUILD_PYTHON_SUPPORT=ON -D WITH_TBB=ON -D BUILD_NEW_PYTHON_SUPPORT=ON -D WITH_XINE=ON -D WITH_V4L=ON D WITH_OPENGL=ON -D WITH_OPENCL=OFF -D CMAKE_INSTALL_PREFIX=/usr/local ..

When complete you should check that your options agree with those displayed. There are many others available. Support for other programming languages may be missed out if they are not already installed. In my case those for Java, if I were to try OpenCV in Java I would need to install the appropriate Java packages using yum, then recompile OpenCV.

Now build and install, this can take a while:

$ make
$ sudo make install

You now need to move the module to anywhere on the Python Path, to find this:

$ python
Python 2.7.5 (default, Nov 12 2013, 16:18:42)
[GCC 4.8.2 20131017 (Red Hat 4.8.2-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> print sys.path
['', '/usr/lib/python2.7/site-packages/PIL-1.1.7-py2.7-linux-x86_64.egg', '/usr/lib64/python27.zip', '/usr/lib64/python2.7', '/usr/lib64/python2.7/plat-linux2', '/usr/lib64/python2.7/lib-tk', '/usr/lib64/python2.7/lib-old', '/usr/lib64/python2.7/lib-dynload', '/usr/lib64/python2.7/site-packages', '/usr/lib64/python2.7/site-packages/PIL', '/usr/lib64/python2.7/site-packages/gst-0.10', '/usr/lib64/python2.7/site-packages/gtk-2.0', '/usr/lib/python2.7/site-packages', '/usr/lib/python2.7/site-packages/setuptools-0.6c11-py2.7.egg-info']

The directory /usr/lib/python2.7/site-packages looks suitable:

$ sudo mv /usr/local/lib/python2.7/site-packages/cv2.so /usr/lib/python2.7/site-packages

And add your new installation to the PYTHONPATH, and add the export to the end of your .bashrc so it survives a reboot:

$ export PYTHONPATH=$PYTHONPATH:/usr/local/lib/python2.7/site-packages
$ echo export PYTHONPATH=$PYTHONPATH:/usr/local/lib/python2.7/site-packages >> ~/.bashrc

Now to test:

$ python
Python 2.7.5 (default, Nov 12 2013, 16:18:42)
[GCC 4.8.2 20131017 (Red Hat 4.8.2-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
>>> print cv2.__version__
2.4.7
>>> i = cv2.SURF()
>>>

Sorted. Now, perhaps, I can get on with what I actually want to do….

Sources:

Converting the Canon GPS log

Having been out and about with the Canon 6D, for once I had remembered to turn on the GPS logging built into the camera. Upon returning home after the seven mile walk I saved the GPS data in the cameras memory to the SD card. Now this LOG file found in the GPS directory of the SD card can be read directly into Google Earth but not into Memory Map for viewing on the far superior OS 1:25,000 Ordnance Survey map, also Memory Map does not read Google Earth’s KML or KMZ files.

So, the LOG file needs converting, but what format is it in? Canon are, it seems, rather unhelpful in revealing the secret but a little googeling found a reference to them using the NMEA-0183 format.

For conversion I found the free, excellent and very comprehensive gpsbabel to do the job (the PDF manual is over 200 pages), for me I convert the GPS files on a linux computer, but they do a Windows version and I assume the method and outcome will be the same. On Debian, install gpsbabel with:
$ sudo apt-get install gpsbabel

The basic use of gpsbabel for converting is as follows:
gpsbabel -i <input format> -f <input file> -o <output format> -F <output file>

As I know Memory Map reads the Garmin GPX format, I chose that as the output:
$ gpsbabel -i nmea -f 13120100.LOG -o gpx -F 13120100.gpx

And the rest is, as they say, Topographic.

walkabout

Python and the Oracle Client

Update 24 Nov 2015: Also, see my post Upgrading the Python Oracle Client for updating from version 11 to 12.

Installing the python cx_Oracle extension module for connecting to Oracle databases on this Fedora 18 workstation turned out to be a bit of a faff by giving an assortment of unhelpful error messages, if you are having the same pain maybe this will help.

You will need two files from the Oracle Database Instant Client download site, the basic client package and the SDK (devel): http://www.oracle.com/technetwork/database/features/instant-client/index-097480.html For Linux choose the correct flavour for your installed operating system: x86 or x86-64 for 64bit operating systems, you will need to register on the site to access the files. For my purposes I got the 11.2 version for x64 RPM files, install them using:
$ sudo rpm -i oracle-instantclient11.2-basic-11.2.0.3.0-1.x86_64.rpm
$ sudo rpm -i oracle-instantclient11.2-devel-11.2.0.3.0-1.x86_64.rpm

You will now need to tell the system where the libraries are:
$ sudo su
# echo /usr/lib/oracle/11.2/client64/lib/ > /etc/ld.so.conf.d/oracle.conf
# ldconfig
# exit

To install cx_Oracle you will need to set some environment variables otherwise you will get an “error: cannot locate an Oracle software installation” message. The easy_install program is found in python-setuptools I have included it in the recipe as a reminder if you have not installed it already.
$ sudo yum install python-setuptools
$ export ORACLE_HOME=/usr/lib/oracle/11.2/client64
$ export LD_LIBRARY_PATH=$ORACLE_HOME/lib:$LD_LIBRARY_PATH
$ export PATH=$ORACLE_HOME/bin:$PATH
$ sudo -E easy_install cx_Oracle

or by downloading the version from http://cx-oracle.sourceforge.net/ and installing it manually, you will still need to set the exports, as above, and do the following for installation:
$ sudo -E python setup.py build
$ sudo -E python setup.py install

The -E on the sudo takes your environment variables, including the three you just set, into your sudo session.

Success:
$ python
Python 2.7.3 (default, Aug 9 2012, 17:23:57)
[GCC 4.7.1 20120720 (Red Hat 4.7.1-5)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import cx_Oracle
>>> exit()

When the the library cannot be seen, you get this error:
$ python
Python 2.7.3 (default, Aug 9 2012, 17:23:57)
[GCC 4.7.1 20120720 (Red Hat 4.7.1-5)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import cx_Oracle
Traceback (most recent call last):
File "", line 1, in
ImportError: libclntsh.so.11.1: cannot open shared object file: No such file or directory
>>> exit()