本笔记本配置N卡1660Ti，Ubuntu18.04。需要用到ros、opencv、tensorflow（依赖于anaconda的版本）、python来实现openpose项目。之前是已有ros-melodic，然后，在anaconda的base环境下安装了python3.6，和conda下载的tensorflow1.13.1，和源码编译的opencv及opencv_contrib版本3。
一、此时环境如下：
conda base虚拟环境：
gedit ~/.bashrc中的变量：
# added by Anaconda3 4.3.0 installer
export PATH="/home/muxi/anaconda3/bin:$PATH"
export PYTHONPATH=/home/muxi/anaconda3/lib/python3.6/site-packages:$PYTHONPATH
#ros
source /opt/ros/melodic/setup.bashpython的环境：
muxi@muxi-Z2-Air:~$ python -c "import sys; print(sys.prefix)"
/home/muxi/anaconda3
muxi@muxi-Z2-Air:~$ which python
/home/muxi/anaconda3/bin/python
muxi@muxi-Z2-Air:~$ which python3
/home/muxi/anaconda3/bin/python3
muxi@muxi-Z2-Air:~$ python --version
Python 3.6.9 :: Anaconda, Inc.
muxi@muxi-Z2-Air:~$ python3 --version
Python 3.6.9 :: Anaconda, Inc.
muxi@muxi-Z2-Air:~$ python -c "from distutils.sysconfig import get_python_inc; print(get_python_inc())"
/home/muxi/anaconda3/include/python3.6m
muxi@muxi-Z2-Air:~$ python -c "from distutils.sysconfig import get_python_lib; print(get_python_lib())"
/home/muxi/anaconda3/lib/python3.6/site-packages
muxi@muxi-Z2-Air:~$ python -c "import distutils.sysconfig as sysconfig; print(sysconfig.get_config_var('LIBDIR'))"
/home/muxi/anaconda3/lib
opencv的环境：
muxi@muxi-Z2-Air:~$ pkg-config opencv --modversion #当前版本
3.2.0
muxi@muxi-Z2-Air:~$ whereis opencv #系统默认在/usr下找OpenCV
opencv: /usr/include/opencv /usr/share/opencv
muxi@muxi-Z2-Air:~$ pkg-config opencv --libs
-lopencv_shape -lopencv_stitching -lopencv_superres -lopencv_videostab -lopencv_aruco -lopencv_bgsegm -lopencv_bioinspired -lopencv_ccalib -lopencv_datasets -lopencv_dpm -lopencv_face -lopencv_freetype -lopencv_fuzzy -lopencv_hdf -lopencv_line_descriptor -lopencv_optflow -lopencv_video -lopencv_plot -lopencv_reg -lopencv_saliency -lopencv_stereo -lopencv_structured_light -lopencv_phase_unwrapping -lopencv_rgbd -lopencv_viz -lopencv_surface_matching -lopencv_text -lopencv_ximgproc -lopencv_calib3d -lopencv_features2d -lopencv_flann -lopencv_xobjdetect -lopencv_objdetect -lopencv_ml -lopencv_xphoto -lopencv_highgui -lopencv_videoio -lopencv_imgcodecs -lopencv_photo -lopencv_imgproc -lopencv_core
muxi@muxi-Z2-Air:~$ python -c "import cv2; print(cv2.__version__)"
3.4.1二、使用opencv的设置：
1、当在外部环境的C++程序里想要调用anaconda里编译的opencv，而不是/usr里的时，cmakelist里要set(OpenCV_DIR ~/opencv_cuda_anacondabase/opencv-3.4.1/build)。
2、当在Python（anaconda环境下）调用opencv，需要注释掉bashrc中source /opt/ros/melodic/setup.bash（因为ROS支持的python与anaconda的python有很复杂的冲突），或者在需要运行的python文件中的最前面添加两行：
import sys
sys.path.remove('/opt/ros/melodic/lib/python2.7/dist-packages')
三、CLion里使用Python项目的设置：
添加conda环境已存在的Python解释器
后来ros崩了，在系统环境里重装ros-melodic时报错：File "/home/muxi/anaconda3/lib/python3.6/site-packages/yaml/__init__.py", line 374
class YAMLObject(metaclass=YAMLObjectMetaclass):
这不是网络引起的超时问题，bashrc中屏蔽export PYTHONPATH，https://www.ncnynl.com/archives/201906/3147.html中sudo aptitude install ros-melodic-desktop-full安装ros解决冲突，roscore成功。
四、ROS和anaconda共存：
在CLion和终端下分别测试cv2和tensorflow，都能在roscore存在下成功，不用remove关于ros的path。
至此，初步搭建起存在cuda、anaconda、py、tensorflow、cv以及ros的C++、PY混合工程环境。
五、创建openpose的ROS工作空间
~./bashrc 里需要有 source opt/ros/melodic/setup.bash
#建立ros的workspace
mkdir -p catkin_openpose/src
cd catkin_openpose/src
catkin_init_workspace #生成catkin_openpose/src/CMakeList.txt
cd ..
catkin_make #编译生成catkin_openpose/build/和catkin_openpose/devel/
#创建功能包
cd src
catkin_create_pkg bagname std_msgs roscpp rospy #后面仨是包的依赖六、rviz
#安装rviz
sudo apt-get install ros-melodic-rviz1、运行 rviz 报错：rviz: error while loading shared libraries: libicui18n.so.58: cannot open shared object file: No such file or directory
执行：
which rviz #输出：/opt/ros/melodic/bin/rviz
ldd /opt/ros/melodic/bin/rviz
建立软链接：
sudo ln -s /home/muxi/anaconda3/lib/libicui18n.so.58 /usr/lib/x86_64-linux-gnu/libicui18n.so.58
sudo ln -s /home/muxi/anaconda3/lib/libicuuc.so.58 /usr/lib/x86_64-linux-gnu/libicuuc.so.58
sudo ln -s /home/muxi/anaconda3/lib/libicudata.so.58 /usr/lib/x86_64-linux-gnu/libicudata.so.58
2、运行 rviz 又报错：
This application failed to start because it could not find or load the Qt platform plugin "xcb"
in "".
Reinstalling the application may fix this problem.
执行：
export QT_DEBUG_PLUGINS=13、再运行 rviz 输出详细报错：
In /home/muxi/Qt5.10.1/5.10.1/gcc_64/plugins/platforms/libqxcb.so:
Plugin uses incompatible Qt library (5.10.1) [release]
"The plugin '/home/muxi/Qt5.10.1/5.10.1/gcc_64/plugins/platforms/libqxcb.so' uses incompatible Qt library. (5.10.1) [release]"
not a plugin
QFactoryLoader::QFactoryLoader() checking directory path "/opt/ros/melodic/lib/rviz/platforms" ...
执行：
ldd /home/muxi/Qt5.10.1/5.10.1/gcc_64/plugins/platforms/libqxcb.so输出依赖正常，未见not found的动态库。
见https://stackoverflow.com/questions/63829991/qt-qpa-plugin-could-not-load-the-qt-platform-plugin-xcb-in-even-though-it ，说The problem was loading Qt in the conda environment.这里应该是说不能用5.10.0。
执行：
qmake -versionQMake version 3.1
Using Qt version 5.9.7 in /home/muxi/anaconda3/lib
执行：
conda list | grep 'qt'
那就用5.9.7吧，运行tf-pose-estimation也是用的5.9.7。
4、bashrc里加：
export QT_PLUGIN_PATH=~/anaconda3/pkgs/qt-5.9.7-h5867ecd_1/pluginsrviz成功。