I create a catkin workspace like:

mkdir -p ~/catkin_ws/src 
cd ~/catkin_ws/src
catkin_init_workspace

After that I create a package:

cd ~/catkin_ws/src 
catkin_create_pkg vp 
cd ~/catkin_ws 
catkin_make

After that I create ~/catkin_ws/src/vp/src/vpnode.cpp:

#include <iostream>
#include "opencv2/core.hpp"
#include "opencv2/videoio.hpp"
#include "opencv2/highgui.hpp"

using namespace std;
using namespace cv;

int main(int argc, char* argv[]) {

    Mat img = imread("/home/andrija/Documents/image.jpg");
    imshow("img", img); // program crashes on this line. when this line is commented the code runs fine, printing out the correct dimensions of image.jpg

    waitKey(2000);
    cout << img.rows << "; " << img.cols << endl;

    return 0;
}

I edit ~/catkin_ws/src/vp/package.xml and add:

<build_depend>opencv2</build_depend>
<run_depend>opencv2</run_depend>

I edit ~/catkin_ws/src/vp/CMakeLists.txt so it looks like this:

cmake_minimum_required(VERSION 2.8.3)
project(vp)

find_package(catkin REQUIRED cv_bridge genmsg image_transport sensor_msgs)

find_package(OpenCV 3 REQUIRED)
catkin_package(
  DEPENDS system_lib
)

include_directories(include ${OpenCV_INCLUDE_DIRS})

add_executable(vpnode src/vpnode.cpp)


target_link_libraries(vpnode
  ${catkin_LIBRARIES}
  ${OpenCV_LIBRARIES}
)

Then I perform:

cd ~/catkin_ws
catkin_make

And the target builds without errors. After that:

source devel/setup.bash

roscore is running the entire time. When I perform rosrun vp vpnode, I get the output:

Segmentation fault (core dumped)

In the comments inside the vpnode.cpp are details about where the crash occurs. I cannot figure out what am I doing wrong. It's like the wrong OpenCV is getting linked, and I don't know why.

Any help?

ADDITIONAL INFO

In CMakeLists.txt, if I don't explicitly put

find_package(OpenCV 3 REQUIRED)

but, instead I do this

find_package(OpenCV REQUIRED)

catkin_make cannot build the target because of the

#include "opencv2/videoio.hpp"

and after I comment that out, catkin_make goes through, but rosrun vp vpnode still reports segmentation fault. So, I'm guessing that there is OpenCV 2 version that is getting mixed up with OpenCV 3.

Im not expert on Linux systems, so I don't know how to fix that. Prior to installing OpenCV3, I deleted everything that had to do with OpenCV2, but now it seems that this isn't true.

EDIT

Thanks for the response, ROSKinect.

I ended up uninstalling both OpenCV and ROS. But, from my recollection, I can respond to your suggestions, since I tried some of them.

Using #include <opencv2/highgui/highgui.hpp> did not help. Neither did getting rid of videoio. I have not tried including cv.hpp. Your suggested code would most likely crash on the line namedWindow in my problem.

I was 95% sure that some config file or something like that was left undeleted from the Opencv2.4, i had installed previously, and my program was trying to access some libraries which weren't available anymore.

After proper uninstallation of both OpenCV and ROS and re-installation, I got rid of this problem. But encounted a new one.

find_package(OpenCV REQUIRED)

add_executable(vpnode
  src/vpnode.cpp
)
target_link_libraries(vpnode
  ${catkin_LIBRARIES}
  ${OpenCV_LIBRARIES}
)

#include <iostream>
#include <opencv/cv.h>
#include <opencv2/highgui/highgui.hpp>

int main(int argc, char* argv[])  {
    cv::Mat img = cv::imread("/home/andrija/Documents/image.jpg"); //Make sure that the image is there

    std::cout << img.rows << "; " << img.cols <<  std::endl;

    cv::namedWindow("img");
    cv::imshow("img", img);

    cv::waitKey(2000);    
    return 0; 
}