关于ros将opencv版本固定“写死”的一些想法

今天主要工作是将ros和zed结合起来，但是发现自己安装了opencv3.1，ros indigo安装的是opencv2.4.8，这就麻烦了，zed支持的是opencv3.1。一开始使用slam2时，报错：

编译出现错误，
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `String':
/usr/local/include/opencv2/core/cvstd.hpp:625: undefined reference to `cv::String::allocate(unsigned long)'
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `main':
/home/car/Desktop/ORB_SLAM2/Examples/ROS/ORB_SLAM2/src/ros_stereo.cc:86: undefined reference to `cv::FileStorage::FileStorage(cv::String const&, int, cv::String const&)'
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `main':
/usr/local/include/opencv2/core/cvstd.hpp:667: undefined reference to `cv::String::deallocate()'
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `~String':
/usr/local/include/opencv2/core/cvstd.hpp:667: undefined reference to `cv::String::deallocate()'
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `cv::FileNode::operator int() const':
/usr/local/include/opencv2/core/persistence.hpp:1180: undefined reference to `cv::read(cv::FileNode const&, int&, int)'
/usr/local/include/opencv2/core/persistence.hpp:1180: undefined reference to `cv::read(cv::FileNode const&, int&, int)'
/usr/local/include/opencv2/core/persistence.hpp:1180: undefined reference to `cv::read(cv::FileNode const&, int&, int)'
/usr/local/include/opencv2/core/persistence.hpp:1180: undefined reference to `cv::read(cv::FileNode const&, int&, int)'
CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: In function `main':
/usr/local/include/opencv2/core/cvstd.hpp:667: undefined reference to `cv::String::deallocate()'
/usr/local/include/opencv2/core/cvstd.hpp:667: undefined reference to `cv::String::deallocate()'

之后：

我将CMakeLists.txt 里面的
find_package(OpenCV 2.4.3 REQUIRED)
改成
find_package(OpenCV 3.1.0 REQUIRED COMPONENTS core highgui imgproc imgcodecs)

好像就可以了,但是，运行出错了
OpenCV Error: Bad argument (Invalid pointer to file storage) in cvGetFileNodeByName, file /build/buildd/opencv-2.4.8+dfsg1/modules/core/src/persistence.cpp, line 740
terminate called after throwing an instance of 'cv::Exception'
  what():  /build/buildd/opencv-2.4.8+dfsg1/modules/core/src/persistence.cpp:740: error: (-5) Invalid pointer to file storage in function cvGetFileNodeByName

OpenCV Error: Bad argument in cvGetFileNodeByName,src/persistence.cpp,  throwing 'cv::Exception'
  what():  /build/buildd/opencv-2.4.8+dfsg1/modules/core/src/persistence.cpp:740: error: (-5) Invalid pointer to file storage in function cvGetFileNodeByName

还是出错，改的心慌慌呀。

网上有个解决方案，转自http://m.blog.csdn.net/article/details?id=52842099

ROS-Kinect cv_bridge的opencv版本兼容问题

发表于2016/10/17 21:32:15  326人阅读

分类： 研究

转自我的博客(http://gongzheng92.net)

---

今天在编译自己的drone_sensor模块的时候（含stereo camera）出现了

/home/joey/pfl/ros/src/drone_sensor/src/stereo_camera/CameraThread.cpp: In member function ‘virtual void CameraThread::loop()’: /home/joey/pfl/ros/src/drone_sensor/src/stereo_camera/CameraThread.cpp:65:47: error: ‘cvDeInterlace’ was not declared in this scope
         cvDeInterlace(img, evenImage, oddImage);
                                               ^
/home/joey/pfl/ros/src/drone_sensor/src/stereo_camera/CameraThread.cpp: In member function ‘void CameraThread::publishImage(image_transport::CameraPublisher&, IplImage*)’: /home/joey/pfl/ros/src/drone_sensor/src/stereo_camera/CameraThread.cpp:85:32: error: no matching function for call to ‘cv::Mat::Mat(IplImage*&)’
     cv::Mat image = cv::Mat(img);

• 1
• 2
• 3
• 4
• 5

的错误（我的catkin_ws即为~/pfl/ros）

我的配置为：

• Ubuntu 16.04 虚拟机
• ROS版本为Kinect，路径为/opt/ros/kinect

通过这条问题里中提到的命令查询opencv的路径与版本还有依赖库：

    pkg-config --modversion opencv
    2.4.9.1

    pkg-config --cflags opencv
    -I/usr/include/opencv

    pkg-config --libs opencv
    #这里有好多.so文件的路径，每个人都不尽一样，略去。

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8

从报错以及网上的资料可以发现造成错误的原因主要是由于ros在Kinect的版本中自带了opencv3（通过ros-kinect-opencv3安装），在目录/opt/ros/kinetic/include/opencv-3.1.0-dev下。cv3与cv2的不兼容造成了错误。具体造成本错误的原因是cvDeInterlace这个函数存在于cvaux.h所引用的opencv2/legacy/legacy.hpp文件中，而在cv3中cvaux并没有引用这个文件（那么在cv3中对应的函数应该如何调用？）

在这个问题中题主遇到了非常相似的情况，然而给出的解决方式：

将cmakelist中
find_package(OpenCV REQUIRED)
改为
find_package(OpenCV 2.4.8 EXACT REQUIRED)
（其中2.4.8为具体版本号，我的情况为2.4.9.1）

并没有用，make还是会去调用ros的cv3.

继续在网上翻找之后发现时cv_bridge的问题。ros kinect的cv_bridge会自动关联cv3。具体在/opt/ros/kinetic/share/cv_bridge/cmake中的cv_bridgeConfig.cmake里可以看到：

第94-96行：
    if(NOT "include;/opt/ros/kinetic/include/opencv-3.1.0-dev/opencv;/opt/ros/kinetic/include/opencv-3.1.0-dev " STREQUAL " ")
      set(cv_bridge_INCLUDE_DIRS "")
      set(_include_dirs "include;/opt/ros/kinetic/include/opencv-3.1.0-dev/opencv;/opt/ros/kinetic/include/opencv-3.1.0-dev")

• 1
• 2
• 3
• 4

以及第112行的所有lib都是cv3的。

因此将所有的这些opencv的路径都改为系统的opencv路径即可解决问题。同时上述的find_package中的信息并没有必要修改。
附上我的修改方案：

#94-96行：
if(NOT "include;/usr/include/opencv;/usr/include " STREQUAL " ")
  set(cv_bridge_INCLUDE_DIRS "")
  set(_include_dirs "include;/usr/include/opencv;/usr/include")
#112行：
set(libraries "cv_bridge;/usr/lib/x86_64-linux-gnu/libopencv_calib3d.s
o;/usr/lib/x86_64-linux-gnu/libopencv_contrib.so;/usr/lib/x86_64-linux-
gnu/libopencv_core.so;/usr/lib/x86_64-linux-gnu/libopencv_features2d.so
;/usr/lib/x86_64-linux-gnu/libopencv_flann.so;/usr/lib/x86_64-linux-gnu
/libopencv_gpu.so;/usr/lib/x86_64-linux-gnu/libopencv_highgui.so;/usr/li
b/x86_64-linux-gnu/libopencv_imgproc.so;/usr/lib/x86_64-linux-gnu/libope
ncv_legacy.so;/usr/lib/x86_64-linux-gnu/libopencv_ml.so;/usr/lib/x86_64-l
inux-gnu/libopencv_objdetect.so;/usr/lib/x86_64-linux-gnu/libopencv_ocl.so
;/usr/lib/x86_64-linux-gnu/libopencv_photo.so;/usr/lib/x86_64-linux-gnu/
libopencv_stitching.so;/usr/lib/x86_64-linux-gnu/libopencv_superres.so;/
usr/lib/x86_64-linux-gnu/libopencv_ts.so;/usr/lib/x86_64-linux-gnu/lib
opencv_video.so;/usr/lib/x86_64-linux-gnu/libopencv_videostab.so")

有了些启发，但是还是没做出来