Install OpenCV + Python under Ubuntu



Install OpenCV + Python under Ubuntu

1         system configuration

(Host) operating system: Win1064 bit, Ubuntu virtual machine under VirtualBox .

Ubuntu version: 16.04.2

Python has been installed before, version: 3.5.2

1.1   Check the Ubuntu version

method one:

~$ uname-a

Method Two:

~$ cat/proc/version

Method three:

~$ lsb_release–a

1.2   View the installation path and version number of Python

method one:

~$whereis python

Method Two:

~$ whichpython

Method 3: Start Python and check the output information

Careful students should be able to find that there are different versions of Python installed on the current Ubuntu host , namely 2.7.12 and 3.5.2 . If you want to start different versions of Python , execute python2 and python3 (or python3. 5 ) Two commands are enough. For example:

~$python3

Start Python-3.5.2 , the prompt is as follows:

Through the detailed information of the list file, you can see the executable files linked by the two commands, as shown in the following figure:

2         Install OpenCV

2.1   Download the OpenCV installation package

Address: https://opencv.org/releases.html (Note: You can directly locate the source code download page of the second part through the link)

The interface is as follows:

I chose to install the latest version ( 3.4.0 ), and click the link address of the corresponding version (the red circle in the figure).

Enter a new page (address: https://github.com/opencv/opencv/releases/tag/3.4.0 ), the interface is as follows:

Choose to download the source code ( zip or tar.gz is acceptable), and I choose to download the tar.gz file, as shown in the red circle in the figure.

Click the link to download the file locally.

2.2   Unzip the file

Order:

~$ tar-zxvf

 

2.3   Dependency package installation

First perform the update operation:

~$ sudoapt-get update

The commands to check whether a package has been installed are:

~$ dpkg-s pkg_name

or

~$ dpkg--get-selections | grep pkg_name

According to the existing strategies on the Internet, the packages that need to be installed / confirmed are:

build-essential cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-devlibswscale-dev

Packages that have not been confirmed to be installed are:

cmake git libgtk2.0-dev libavcodec-dev libavformat-devlibswscale-dev

Execute the following command to perform the installation:

~$ sudoapt-get install cmake git libgtk2.0-dev libavcodec-dev libavformat-devlibswscale-dev

2.4   Compile and install OpenCV

Enter the decompressed file directory, and execute the following command:

~$ cd ~/opencv-3.4.0

~$ mkdir release

~$ cd release

~$ cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local..

~$ make

~$ sudo make install

 

The make command takes about 26 minutes! (It is estimated that it is due to the virtual machine)

 

So far, OpenCV can be used through c/c++ ( OpenCV 's implementation language) .

In order to use OpenCV in Python , some additional work is required ( after testing, it is confirmed that there is no need to perform the additional work mentioned in the existing network guides ), continue...

3         Test whether the installation is successful

3.1   Install OpenCV package for Python

As in Section 1.2 above, there are multiple Python versions installed on this machine , but after inspection, the three pip commands under /usr/local/bin (respectively pip, pip3 and pip3.5 ) are exactly the same (can be verified by file comparison), And both point to python3.5 . The first line of the three files is #!/usr/bin/python3 as shown below:

I suddenly recalled that during the installation process, in the step sudo make install , there was a message indicating what link has been created for python (including two versions of Python ), and I forgot to take a screenshot!

So it may not be necessary to install the python-opencv package mentioned in various strategies on the Internet .

So test it directly.

3.2   Installation test

Test one:

Start python , I start python3 here , and execute the code shown below, success! The interface and information are as follows:

 

Test two:

Type the following code in the current directory and save it as a file: test01.py

Another thing to do is to create an images directory in the current directory, and copy a picture ( test01.png ) to this directory. Of course, the storage path, format, and file name of the image file are optional, as long as they match the code That's it.

Execute the Python file, a pop-up window will display the specified picture, and you're done!