ORB-SLAM2 | Prometheus_px4 | OpenCV 3.4.9
Reference to ORB-SLAM2 | GTK+ 2.x symbols detected. Using GTK+ 2.x and 3 in the same process is not supported
https://zhuanlan.zhihu.com/p/400316912
1、环境描述
操作系统：ubuntu18.04.5 LTS （AMD64）
相关软件：ORB-SLAM2、ROS-melodic（OpenCV：3.2.0 installed）、OpenCV：3.4.9（手动编译，编译时CMake添加参数：-DWITH_GTK_2_X=ON）、Prometheuse PX4 (v1)
ROS使用包（与错误相关）：opencv_bridge、image_geometry
编译指令：catkin_make
注：实际这里的 ORB-SLAM 2 是经过修改且加了很多其他功能的。主要用于 prometheus_px4 中的场景仿真。
2、命令行错误信息
在屏幕上出现的如下的错误提示：
/usr/bin/ld: warning: libopencv_imgcodecs.so.3.2, needed by /opt/ros/melodic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_core.so.3.2, needed by /opt/ros/melodic/lib/libcv_bridge.so, may conflict with libopencv_core.so.3.4
/usr/bin/ld: warning: libopencv_imgproc.so.3.4, needed by /usr/lib/gcc/x86_64-linux-gnu/8/../../../x86_64-linux-gnu/libopencv_imgcodecs.so, may conflict with libopencv_imgproc.so.3.2
[ 8%] Built target web_cam
[ 10%] Linking CXX shared library /home/user/Prometheus/devel/lib/libellipselib.so
[ 10%] Built target ellipselib
Makefile:159: recipe for target 'all' failed
make: *** [all] Error 2
Invoking "make -j4 -l4" failed
[INFO] SUBMODULE object_detection_landing NOT EXIST, Skip it!
[INFO] SUBMODULE object_detection_yolov5openvino NOT EXIST, Skip it!
[ 15%] Linking CXX executable /home/user/Prometheus/devel/lib/prometheus_detection/landpad_det
/usr/bin/ld: warning: libopencv_imgcodecs.so.3.2, needed by /opt/ros/melodic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_core.so.3.2, needed by /opt/ros/melodic/lib/libcv_bridge.so, may conflict with libopencv_core.so.3.4
3、排查过程
该问题陆续困扰我很久，简要说一下正确解决路线：
3.1、ORB-SLAM 相关问题
从提示中可猜测到，ORB-SLAM在编译过程中使用了多个不同版本的OpenCV库，因此造成了冲突。
要排查到底是具体哪些库的使用之相关，则需要引入一个查看动态库依赖的命令：
ldd #（list dynamic dependencies）
这个命令的作用是列出程序或动态库的依赖关系。
例如：查看 libslam_semantic_nav_ros.so 库文件运行所依赖的库:
~/slam_ws/src/orbslam2/lib$ ldd libslam_semantic_nav_ros.so | grep libopencv
libopencv_calib3d.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_calib3d.so.3.2 (0x00007fce94e1c000)
libopencv_core.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_core.so.3.2 (0x00007fce949e1000)
libopencv_features2d.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_features2d.so.3.2 (0x00007fce9472c000)
libopencv_highgui.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_highgui.so.3.2 (0x00007fce94520000)
libopencv_imgproc.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_imgproc.so.3.2 (0x00007fce93fcf000)
libopencv_video.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_video.so.3.2 (0x00007fce93d6c000)
libopencv_text.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_text.so.3.2 (0x00007fce93a4d000)
libopencv_core.so.3.4 => /usr/local/lib/libopencv_core.so.3.4 (0x00007fce8fef8000) # 版本与前面的不同。
libopencv_imgcodecs.so.3.4 => /usr/local/lib/libopencv_imgcodecs.so.3.4 (0x00007fce8fa9f000) # 版本与前面的不同。
libopencv_ml.so.3.4 => /usr/local/lib/libopencv_ml.so.3.4 (0x00007fce8f3e6000)
libopencv_flann.so.3.4 => /usr/local/lib/libopencv_flann.so.3.4 (0x00007fce8d785000)
libopencv_flann.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_flann.so.3.2 (0x00007fce8c287000)
libopencv_imgcodecs.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_imgcodecs.so.3.2 (0x00007fce8a13f000)
libopencv_ml.so.3.2 => /usr/lib/x86_64-linux-gnu/libopencv_ml.so.3.2 (0x00007fce896df000)
libopencv_imgproc.so.3.4 => /usr/local/lib/libopencv_imgproc.so.3.4 (0x00007fce82950000)
上面列出的信息：
#第一列：程序需要依赖什么库
#第二列: 程序所需库在系统中的路径位置
#第三列：即括号中的数，为库加载的开始地址
#如果需要某个库，但程序在系统找不到对应的库，会出现第5行的not found信息
3.2、通过 ldd 命令排查库依赖
一般而言 Prometheus 和 ORB-SLAM2 本身用的库：
orbslam2/lib/liborbslam2.so
orbslam2/Thirdparty/DBoW2/lib/libDBoW2.so
orbslam2/Thirdparty/g2o/lib/libg2o.so
opencv、eigen3等 ORB-SLAM2 所需的库
ros 功能包的相关库等。
首先，通过使用 ldd liborbslam2.so | grep opencv 命令发现 liborbslam2.so 确实链接了 opencv-3.2.0 和 opencv-3.4.9 两个版本的库，输出如下：
~/Prometheus/devel/lib$ ldd libkcftrackerlib.so | grep opencv
libopencv_core.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_core.so.3.4 (0x00007f6db569a000)
libopencv_imgproc.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_imgproc.so.3.2 (0x00007f6db3e87000)
libopencv_core.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_core.so.3.2 (0x00007f6db199d000)
继续查看 prometheus_detection 下的 stereo_detection_saving_gazebo_samples 的库依赖
~/Prometheus/devel/lib/prometheus_detection$ ldd stereo_detection_saving_gazebo_samples | grep opencv
libopencv_core.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_core.so.3.4 (0x00007efcba027000)
libopencv_imgcodecs.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_imgcodecs.so.3.4 (0x00007efcb96c9000)
libopencv_highgui.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_highgui.so.3.4 (0x00007efcb88c0000)
libopencv_imgcodecs.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_imgcodecs.so.3.2 (0x00007efcb6cb8000)
libopencv_imgproc.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_imgproc.so.3.2 (0x00007efcb519d000)
libopencv_core.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_core.so.3.2 (0x00007efcb42af000)
libopencv_imgproc.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_imgproc.so.3.4 (0x00007efcb1fd5000)
再次确认一下 prometheus_detection 中 web_cam 的库依赖情况。
~/Prometheus/devel/lib/prometheus_detection$ ldd web_cam | grep opencv
libopencv_core.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_core.so.3.4 (0x00007fb6482a3000)
libopencv_imgproc.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_imgproc.so.3.2 (0x00007fb646a90000)
libopencv_imgcodecs.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_imgcodecs.so.3.4 (0x00007fb646132000)
libopencv_videoio.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_videoio.so.3.4 (0x00007fb6454e7000)
libopencv_imgcodecs.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_imgcodecs.so.3.2 (0x00007fb6436da000)
libopencv_core.so.3.2 => /home/npuyin/ThirdParty/OpenCV-3.2.0_install/lib/libopencv_core.so.3.2 (0x00007fb6424e4000)
libopencv_imgproc.so.3.4 => /usr/local/opencv-3.4.9/lib/libopencv_imgproc.so.3.4 (0x00007fb63ffe0000)
3.3 、排查结果
上述排查已经基本确定了是 opencv 等库依赖出现问题。具体来源是由于 ROS 系统安装的时候自带了OpenCV 3.2.0 （lib 文件安装位置位于 /usr/lib/x86_64-linux-gnu/），而我自用的OpenCV 3.4.9 （安装在/usr/local/lib/）。在编译链接orbslam工程的时候，由于不恰当的配置造成了一部分文件引用了OpenCV 3.2.0，而一部分文件引用了OpenCV 3.4.9 。
确认位置：
$ ls /usr/lib/x86_64-linux-gnu/ | grep opencv
libopencv_aruco.a
libopencv_aruco.so
libopencv_aruco.so.3.2
libopencv_aruco.so.3.2.0
libopencv_bgsegm.a
libopencv_bgsegm.so
libopencv_bgsegm.so.3.2
libopencv_bgsegm.so.3.2.0
libopencv_bioinspired.a
libopencv_bioinspired.so
libopencv_bioinspired.so.3.2
libopencv_bioinspired.so.3.2.0
libopencv_calib3d.a
libopencv_calib3d.so
libopencv_calib3d.so.3.2
libopencv_calib3d.so.3.2.0
libopencv_ccalib.a
libopencv_ccalib.so
libopencv_ccalib.so.3.2
libopencv_ccalib.so.3.2.0
libopencv_core.a
libopencv_core.a_backup
libopencv_core.so
libopencv_core.so.3.2
libopencv_core.so.3.2.0
libopencv_core.so.3.2.0_backup
libopencv_core.so.3.4
libopencv_core.so.3.4.9
……
3.4、解决问题的思路
问题转到：查找ROS自带 OpenCV 库与 自行安装 OpenCV 冲突问题。
单纯 修改/opt/ros/melodic/share/cv_bridge/cmake下*Config.cmake文件版本库，或者更改 CmakeLists.txt 中的 find_package() 并不能完全解决问题。
cv_bridgeConfig.cmake
cv_bridgeConfig-version.cmake
cv_bridge-extras.cmake
编译 orbslam 工程的 CmakeLists.txt 中使用的 ROS 功能包 opencv_bridge 和 image_geometry 的库依赖了OpenCV 3.2.0
4、解决办法
参考相关文献给出的两种方法：
4.1、单独卸载这俩ROS功能包（没有验证）
适用于源码编译安装的 ROS 朋友：去 github 下载相应的包，放到 ros 工作空间的src文件夹下，配置他们的CmakeLists.txt 文件中相应的编译路径，使他们编译的时候使用自己安装的 OpenCV 3.4.15 库。
4.2、直接卸载ros自带的OpenCV（我使用的）
思路简述：
首先、卸载该 OpenCV（即3.2.0版本） 会使得所有依赖此opencv-3.2.0 的 ROS 功能包全部卸载。卸载完成后，去 github 下载工程所需要的缺失的功能包（opencv_bridge 和 image_geometry），放到 ros 工作空间的 src 文件夹下，配置他们的 CmakeLists.txt 文件，使他们编译的时候使用自己安装的 OpenCV 库。
步骤：
1、使用 dpkg --get-selections | grep opencv 指令查找所有opencv相关包
输出可能如下所示：
~$ dpkg --get-selections | grep opencv
libopencv-calib3d-dev:amd64 install
libopencv-calib3d3.2:amd64 install
libopencv-contrib-dev:amd64 install
libopencv-contrib3.2:amd64 install
libopencv-core-dev:amd64 install
libopencv-core3.2:amd64 install
libopencv-dev install
libopencv-features2d-dev:amd64 install
libopencv-features2d3.2:amd64 install
libopencv-flann-dev:amd64 install
libopencv-flann3.2:amd64 install
libopencv-highgui-dev:amd64 install
libopencv-highgui3.2:amd64 install
libopencv-imgcodecs-dev:amd64 install
libopencv-imgcodecs3.2:amd64 install
libopencv-imgproc-dev:amd64 install
libopencv-imgproc3.2:amd64 install
libopencv-ml-dev:amd64 install
libopencv-ml3.2:amd64 install
libopencv-objdetect-dev:amd64 install
libopencv-objdetect3.2:amd64 install
libopencv-photo-dev:amd64 install
libopencv-photo3.2:amd64 install
libopencv-shape-dev:amd64 install
libopencv-shape3.2:amd64 install
libopencv-stitching-dev:amd64 install
libopencv-stitching3.2:amd64 install
libopencv-superres-dev:amd64 install
libopencv-superres3.2:amd64 install
libopencv-ts-dev:amd64 install
libopencv-video-dev:amd64 install
libopencv-video3.2:amd64 install
libopencv-videoio-dev:amd64 install
libopencv-videoio3.2:amd64 install
libopencv-videostab-dev:amd64 install
libopencv-videostab3.2:amd64 install
libopencv-viz-dev:amd64 install
libopencv-viz3.2:amd64 install
libopencv3.2-java install
libopencv3.2-jni install
opencv-data install
python-opencv install
ros-melodic-vision-opencv install
2、从输出中看到 libopencv-core3.2:amd64 install
使用命令 sudo apt-get remove libopencv-core3.2 卸载opencv3.2.0，注意不要添加 purge 参数
3、输完指令后会出现如下类似输出
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following packages were automatically installed and are no longer required:
ros-melodic-compressed-depth-image-transport ros-melodic-compressed-image-transport ros-melodic-cv-bridge ros-melodic-depth-image-proc
ros-melodic-desktop ros-melodic-desktop-full ros-melodic-gazebo-plugins ros-melodic-gazebo-ros-pkgs ros-melodic-image-geometry
ros-melodic-image-pipeline ros-melodic-image-proc ros-melodic-image-publisher ros-melodic-image-rotate ros-melodic-image-transport-plugins
ros-melodic-image-view ros-melodic-perception ros-melodic-rqt-common-plugins ros-melodic-rqt-image-view ros-melodic-simulators
ros-melodic-stereo-image-proc ros-melodic-theora-image-transport ros-melodic-vision-opencv ros-melodic-viz
0 upgraded, 0 newly installed, 5 to remove and 0 not upgraded.
After this operation, 3.8 MB disk space will be freed.
Do you want to continue? [Y/n]
卸载完成后，新开一个终端，输入 roscore，若成功运行，则基本无问题。
4、下载 ros 对应功能包
网址：https://github.com/ros-perception/vision_opencv
将 opencv_bridge 和 image_geometry ，放到 ros 工作空间的 src 文件夹下。
使用 pkg-config --modversion opencv 查看当前 OpenCV 版本，输出 3.4.9 则无问题。
~$ pkg-config --modversion opencv
3.4.9
5、后续
删除ROS工作空间之前的build和devel文件，重新编译（catkin_make）整个工程。
6、附
一些相关问题网页链接：
1 cv_bridge依赖opencv版本的问题
2 OpenCV error when trying to run ROS examples
3 OpenCV error when running ros examples in Ubuntu + ROS
4 关于解决libopencv_core.so.2.4, needed by … 的问题
5 解决Ubuntu环境下ROS 中的cv_bridge和自装OpenCV兼容性笔记
6 ubuntu下多版本opencv切換
7 ·Ubuntu 18.04 ROS-melodic 源码安装指定版本opencv 和 opencv_contrib 使用nonfree模块surf算法
8. 在Ubuntu中通过CV_bridge更改OpenCV版本
https://blog.csdn.net/weixin_48110541/article/details/125176144
7、后记
安装后，Terminal 启动时的相关提示
GAZEBO_PLUGIN_PATH :/home/user/Prometheus/devel/lib:/home/user/prometheus_px4/build/amovlab_sitl_default/build_gazebo
GAZEBO_MODEL_PATH :/home/user/Prometheus/Simulator/gazebo_simulator/models:/home/user/Prometheus/Simulator/gazebo_simulator/amov_models:/home/user/prometheus_px4/Tools/sitl_gazebo/models
LD_LIBRARY_PATH /home/user/Prometheus/devel/lib:/opt/ros/melodic/lib:/usr/local/opencv-3.4.9/lib:/usr/local/cuda-11.1/lib64:/usr/local/cuda/lib64:/home/user/prometheus_px4/build/amovlab_sitl_default/build_gazebo
可以在 /home/npuyin/prometheus_px4/Tools/setup_gazebo.bash 中注释下面三行来关闭显示！
# echo -e "GAZEBO_PLUGIN_PATH $GAZEBO_PLUGIN_PATH"
# echo -e "GAZEBO_MODEL_PATH $GAZEBO_MODEL_PATH"
# echo -e "LD_LIBRARY_PATH $LD_LIBRARY_PATH"
8、临时记录
如果有某个功能包或者软件的安装需要用新的覆盖，例如 php7.2 的相关模块重新安装。最好的办法是到：
cd /var/lib/dpkg/info
ls | grep php
sudo mkdir PHP7.2_BK
sudo mv ./php7.2* ./PHP7.2_BK
cd ~/
sudo apt install php7.2-*
即从注册表中删除该模块，然后再重新安装。
9、忠告
不要在同一个主控上安装多个版本opencv，那是自找麻烦，涉及到各种配置，配置不对就导致程序无法运行，调试配置花费大量精力！
来源：
cv_bridge与opencv版本不一致导致程序编译错误及无法运行程序问题的解决方案（附带ORB_SLAM3案例）
https://blog.csdn.net/qq_18276949/article/details/114820705
1、问题描述
换做其他开发环境遇到的问题也是类似的，解决方案也是类似的，因此本解决方案是通用的。
=================
开发环境
嵌入式主控：nvidia xavier
操作系统：ubuntu 18.04, ROS melodic
（1）ros melodic 默认安装 opencv3.2
（2）使用的 zed 相机需要在 xavier 上安装了CUDA10.0，并默认安装了opencv4.1.1
（3）系统源码安装了另一个opencv版本 opencv3.4.9
总的来说，问题中的主控 xavier 上总共有3个版本的opencv：3.2, 4.1.1, 3.4.9
=================
程序配置
主控中 ROS 节点程序因为要使用 zed 相机，对应的 cmakelists.txt 中编译用的是find_package(OpenCV 4 REQUIRED)，如果用3就导致程序根本编译不通过。
这就导致用户程序调用的 opencv 库版本4.1.1与 ROS 默认安装的 cv_bridge 调用的opencv-3.2.0不一致问题！
一句话总结：ros-melodic 默认安装 cv_bridge-1.13.0 以及 opencv-3.2.0，而用户程序中用到了 cv_bridge 且设置调用的 opencv 版本为4.1.1。
===================================================
2、问题造成的后果
版本不一致的 cv_bridge 与 opencv 问题会导致很棘手的一个问题——用户代码编译成功，但运行报错的问题。
例如：
（1） 跑 ORB_SLAM2 和/或 ORB_SLAM3 失败
（2） 跑基于 opencv 库函数的用户程序失败
等等
——以下展示了ORB_SLAM3编译其ROS例程结果，只有几个节点编译成功，其他节点报出由于opencv版本冲突问题导致编译失败错误。
3、解决方案
总体思路：使用高版本的OpenCV对应的cv_bridge。例如：由于用户程序或 ORB_SLAM3 只能使用 opencv4 才能编译通过，因此解决方案只能是安装使用 opencv4 的 cv_bridge。
另外：ROS noetic 默认安装 cv_bridge1.15.0 以及 opencv-4.2.0。因此，需要下载1.15.0 版本的 cv_bridge 源码，将其放到用户自建的 ROS 工作空间中使用。
然后：catkin_make 编译，之后将用户 ROS 节点程序设置为优先链接用户安装的这个版本 cv_bridge（而不是ROS默认安装的那个）。
最后：在~/.bashrc文件中将用户 ROS 工作空间包的 setup.bash 文件在最后 source就能够覆盖 ROS 系统默认的 cv_bridge了，并在 ROS 包对应的 camkelists.txt 中更新为 find_package(OpenCV 4 REQUIRED) 。
步骤总结：
1. 下载1.15.0版本的cv_bridge源码，catkin_make编译
2. ~/.bashrc中source上步安装的cv_bridge包路径
3. camkelists.txt中更新find_package(OpenCV 4 REQUIRED)
说明：
（1）cv_bridge包包含在vision_opencv包中，因此，建议将vision_opencv包一起下载下来整体编译安装
（2）注意，对应下载源码时选择
noetic ==> vision_opencv-1.15.0
melodic ==> vision_opencv-1.13.0
不同版本 opencv 的安装路径，可以在对应的 ${INSTALL_PATH}/lib/pkgconfig/opencv.pc 中查看到！