配置VINS_Fisheye（vins_fusion鱼眼相机版）

GitHub - xuhao1/VINS-Fisheye: Fisheye version of VINS-Fusion

1.搭建环境

1.1安装CUDA-----电脑已经安装（太好了，不用自己弄）

查看是否安装好

nvcc -V
#输出
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Sun_Oct_23_22:16:07_PDT_2022
Cuda compilation tools, release 11.4, V11.4.315
Build cuda_11.4.r11.4/compiler.31964100_0

1.2安装OpenCV

重新安装OpenCV 3.4.0（不能安装3.4.0，要安装3.4.8以上版本，我开始装错了），因为需要带CUDA版本的，参考Ubuntu18.04安装cuda10以及cuda版本的opencv3.4.13_opencv cuda版本_MWY123_的博客-CSDN博客

1.2.1下载源码

Release OpenCV 3.4.0 · opencv/opencv · GitHub

其他版本可查看：​​​​​​Releases - OpenCV​​​​​

下载好的对应版本的opencv_contrib-3.3.0 置于 opencv-3.3.0文件夹下

1.2.2编译

mkdir build
cd build
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local/opencv3410 -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib-3.4.10/modules/ -D CUDA_CUDA_LIBRARY=/usr/local/cuda-11.4/targets/aarch64-linux/lib/stubs/libcuda.so -D CUDA_ARCH_BIN=8.7 -D CUDA_ARCH_PTX="" -D WITH_CUDA=ON -D WITH_TBB=ON -D BUILD_NEW_PYTHON_SUPPORT=ON -D WITH_V4L=ON -D INSTALL_C_EXAMPLES=ON -D INSTALL_PYTHON_EXAMPLES=ON -D BUILD_EXAMPLES=ON -D WITH_QT=ON -D WITH_OPENGL=ON -D ENABLE_FAST_MATH=1 -D CUDA_FAST_MATH=1 -D WITH_CUBLAS=1 -D WITH_NVCUVID=ON -D BUILD_opencv_cudacodec=OFF ..
make -j4
sudo make install

1.2.3出现的问题

1.2.3.1

CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_nppicom_LIBRARY (ADVANCED)
    linked by target "opencv_cudev" in directory /home/amov/opencv-3.4.0/modules/cudev
    linked by target "opencv_cudev" in directory /home/amov/opencv-3.4.0/modules/cudev
    linked by target "opencv_test_cudev" in directory /home/amov/opencv-3.4.0/modules/cudev/test
    linked by target "opencv_core" in directory /home/amov/opencv-3.4.0/modules/core
    linked by target "opencv_core" in directory /home/amov/opencv-3.4.0/modules/core
    linked by target "opencv_test_core" in directory /home/amov/opencv-3.4.0/modules/core

解决办法：去github对应的版本分支的cmake文件夹中下载FindCUDA.cmake、OpenCVDetectCUDA.cmake替换掉自己的（是不是我一开始下的代码版本就弄错了？）

make时发生的问题

1.2.3.2

/home/amov/opencv-3.4.0/opencv_contrib-3.4/modules/reg/include/opencv2/reg/mapaffine.hpp:72:65: error: expected ‘;’ at end of member declaration
   72 |  CV_WRAP void inverseWarp(InputArray img1, OutputArray img2) const CV_OVERRIDE;
      |                                                              ^~~~~
      |                                                                   ;
/home/amov/opencv-3.4.0/opencv_contrib-3.4/modules/reg/include/opencv2/reg/mapaffine.hpp:72:71: error: ‘CV_OVERRIDE’ does not name a type
   72 |  CV_WRAP void inverseWarp(InputArray img1, OutputArray img2) const CV_OVERRIDE;
      |                                                                    ^~~~~~~~~~~

解决办法：opencv和opencv_contrib版本不一致....下个一样的就行了

1.2.3.3

#error Please include the appropriate gl headers before including cuda_gl_interop.h

解决办法：Cannot Install OpenCV 3.3.1 on JetPack 4.2.3 - Jetson & Embedded Systems / Jetson Nano - NVIDIA Developer Forums

cp /usr/local/cuda/targets/aarch64-linux/include/cuda_gl_interop.h /usr/local/cuda/targets/aarch64-linux/include/cuda_gl_interop.h.org
perl -0 -i -pe 's|#else /\* __APPLE__ \*/[\s\S]*#endif /\* __APPLE__ \*/|#else /* __APPLE__ */\n\n#include <GL/gl.h>\n\n#endif /* __APPLE__ */|gim' /usr/local/cuda/targets/aarch64-linux/include/cuda_gl_interop.h

 1.2.3.4

ollect2: error: ld returned 1 exit status
make[2]: *** [samples/gpu/CMakeFiles/example_gpu_surf_keypoint_matcher.dir/build.make:114：bin/gpu-example-surf_keypoint_matcher] 错误 1
make[1]: *** [CMakeFiles/Makefile2:23665：samples/gpu/CMakeFiles/example_gpu_surf_keypoint_matcher.dir/all] 错误 2

解决办法：把cmake的build_example设置为off（不知道有没有影响）

1.2.3.5

安装3.4.10时遇到

fatal error: boostdesc_bgm.i: 没有那个文件或目录

参考致命错误：boostdesc_bgm.i：没有这样的文件或目录 ·期刊 #1301 ·opencv/opencv_contrib ·GitHub的

1.2.4配置环境变量

sudo gedit /etc/ld.so.conf.d/opencv.conf
#  在文件末尾添加下面语句
/usr/local/opencv3410/lib
# 执行生效
sudo ldconfig
 
gedit ./.bashrc
export PKG_CONFIG_PATH=/usr/local/opencv3410/lib/pkgconfig
export LD_LIBRARY_PATH=/usr/local/opencv3410/lib
source ~/.bashrc
sudo updatedb

终于装好了，浪费我两天

1.3安装高版本的cmake（libSGM需要）

file /lib/systemd/systemd

 查看系统架构（一般来说都是x86_64）

#下载对应版本的camke
wget https://cmake.org/files/v3.12/cmake-3.12.0-Linux-x86_64.tar.gz
#解压
tar -xzvf cmake-3.12.0-Linux-x86_64.tar.gz
#移动文件位置
sudo mv cmake-3.12.0-Linux-x86_64 /opt/cmake-3.12.0
#软链接
sudo ln -sf /opt/cmake-3.12.0/bin/*  /usr/bin/
#查看版本
cmake --version

1.3安装libSGM

下载

编译

2.跑算法

下载源码，创建工作空间，编译

2.1遇到问题

2.1.1

opencv版本冲突，库链接到ros自带的opencv4.2上去了

/usr/bin/ld: warning: libopencv_core.so.3.4, needed by /home/amov/catkin_vins_fusion/devel_isolated/camera_models/lib/libcamera_models.so, may conflict with libopencv_core.so.4.2
/usr/bin/ld: /home/amov/catkin_vinsfisheye/devel/.private/vins/lib/libstereo_depth.so: undefined reference to `cv::_InputArray::_InputArray(cv::MatExpr const&)'

参考ROS中catkin_make的OpenCV冲突的解决（踩坑小记，报错分析）_/usr/bin/ld: warning: libopencv_imgcodecs.so.3.2, _MendozaG的博客-CSDN博客修改了cv_bridgeConfig.cmake文件

2.1.2

make[2]: *** 没有规则可制作目标“/usr/local/opencv3410/lib/libopencv_dnn_superres.so.3.4.10”，由“/home/amov/catkin_vinsfisheye/devel/.private/vins/lib/libstereo_depth.so” 需求。 停止。

离谱，OpenCV3.4.X版本没有这个模块，OpenCV4.X才有，那为什么作者可以在3.4.1模块上测试