Ubuntu 16.04 CUDA9.0 安装ORB_SLAM（上）（原创）

3 安装Pangolin，用于可视化和用户接口
3.1 安装依赖库
3.2 下载Pangolin并配置环境

cd ~/catkin_ws/src

git clone https://github.com/stevenlovegrove/Pangolin

cd ..

catkin_make

source ~/catkin-ws/devel/setup.bash

此步骤执行到catkin_make时，报错：（暂时未解决，先执行下一步）

-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- ~~  traversing 1 packages in topological order:
-- ~~  - pangolin (plain cmake)
-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
CMake Error at /opt/ros/kinetic/share/catkin/cmake/catkin_workspace.cmake:95 (message):
  This workspace contains non-catkin packages in it, and catkin cannot build
  a non-homogeneous workspace without isolation.  Try the
  'catkin_make_isolated' command instead.
Call Stack (most recent call first):
  CMakeLists.txt:63 (catkin_workspace)


-- Configuring incomplete, errors occurred!
See also "/home/l/catkin_ws/build/CMakeFiles/CMakeOutput.log".
See also "/home/l/catkin_ws/build/CMakeFiles/CMakeError.log".

3.3 编译并安装Pangolin

cd Pangolin

mkdir build

cd build

cmake ..

make -j8

sudo make install

4 安装OpenCV
4.1 官网下载OpenCV
建议版本2.4.13，3以上的版本未测试。
4.2 安装编译工具

sudo apt-get install build-essential -y

4.3 安装依赖包

sudo apt-get install libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev -y

4.4 安装可选包

sudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev -y 
sudo apt-get install libgtk2.0-dev -y 
sudo apt-get install pkg-config -y

4.5 进入安装目录 ~/catkin_ws/src ，OpenCV解压至当前目录下。

wget https://github.com/Itseez/opencv/archive/2.4.13.zip

unzip 2.4.13.zi

4.6 编译安装OpenCV源码

cd opencv-2.4.13/

mkdir build

cd build

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local .. 

make

sudo make install

CMAKE后的信息为：

-- -----------------------------------------------------------------
CMake Warning at cmake/OpenCVPackaging.cmake:23 (message):
  CPACK_PACKAGE_VERSION does not match version provided by version.hpp
  header!
Call Stack (most recent call first):
  CMakeLists.txt:1105 (include)


-- Configuring done
-- Generating done
-- Build files have been written to: /home/l/catkin_ws/src/opencv-2.4.13/build

MAKE后报错：（此处问题一直解决不了…）

nvcc fatal   : Unsupported gpu architecture 'compute_20'
CMake Error at cuda_compile_generated_matrix_operations.cu.o.cmake:208 (message):
  Error generating
  /home/l/catkin_ws/src/opencv-2.4.13/build/modules/core/CMakeFiles/cuda_compile.dir/__/dynamicuda/src/cuda/./cuda_compile_generated_matrix_operations.cu.o


modules/core/CMakeFiles/opencv_core.dir/build.make:63: recipe for target 'modules/core/CMakeFiles/cuda_compile.dir/__/dynamicuda/src/cuda/cuda_compile_generated_matrix_operations.cu.o' failed
make[2]: *** [modules/core/CMakeFiles/cuda_compile.dir/__/dynamicuda/src/cuda/cuda_compile_generated_matrix_operations.cu.o] Error 1
CMakeFiles/Makefile2:890: recipe for target 'modules/core/CMakeFiles/opencv_core.dir/all' failed
make[1]: *** [modules/core/CMakeFiles/opencv_core.dir/all] Error 2
Makefile:160: recipe for target 'all' failed
make: *** [all] Error 2

解决过程：
方法一：

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D CUDA_GENERATION=Kepler ..

输出：

CMake Warning at cmake/OpenCVPackaging.cmake:23 (message):
  CPACK_PACKAGE_VERSION does not match version provided by version.hpp
  header!
Call Stack (most recent call first):
  CMakeLists.txt:1105 (include)


-- Configuring done
-- Generating done
-- Build files have been written to: /home/l/catkin_ws/src/opencv-2.4.13/build

MAKE后报错：

[ 98%] Building CXX object modules/gpu/CMakeFiles/opencv_test_gpu.dir/test/nvidia/TestTranspose.cpp.o
[ 98%] Building CXX object modules/gpu/CMakeFiles/opencv_test_gpu.dir/test/nvidia/TestIntegralImage.cpp.o
[ 98%] Linking CXX executable ../../bin/opencv_perf_gpu
[ 98%] Built target opencv_perf_gpu
[ 98%] Linking CXX executable ../../bin/opencv_test_gpu
[ 98%] Built target opencv_test_gpu
Makefile:160: recipe for target 'all' failed
make: *** [all] Error 2


/home/l/catkin_ws/src/opencv-2.4.13/modules/contrib/src/rgbdodometry.cpp:65:47: **fatal error:** unsupported/Eigen/MatrixFunctions: No such file or directory
compilation terminated.
modules/contrib/CMakeFiles/opencv_contrib.dir/build.make:374: recipe for target 'modules/contrib/CMakeFiles/opencv_contrib.dir/src/rgbdodometry.cpp.o' failed
make[2]: *** [modules/contrib/CMakeFiles/opencv_contrib.dir/src/rgbdodometry.cpp.o] Error 1

报错eigen相关，所以就先安装eigen了：
5 安装Eigen3
5.1 最低要求版本为3.1.0。在http://eigen.tuxfamily.org 下载 Eigen3.2.10 压缩文件，并解压。
下载并解压 Eigen3.2.10 到~/catkin_ws/src

 1.    wget https://bitbucket.org/eigen/eigen/get/3.2.10.tar.bz2
 2.    tar -xjf 3.2.10.tar.bz2

编译安装：

cd eigen-3.2.10/

mkdir build

cd build

cmake ..

make

sudo make install

Eigen顺利安装完成。

此时，在回到opencv的build目录下，重新make,
此时，比上次make顺利很多，没有那么多warning和note，但是，仍然报错：

[ 87%] Building CXX object modules/contrib/CMakeFiles/opencv_contrib.dir/src/rgbdodometry.cpp.o
/home/l/catkin_ws/src/opencv-2.4.13/modules/contrib/src/rgbdodometry.cpp:65:47: fatal error: unsupported/Eigen/MatrixFunctions: No such file or directory
compilation terminated.
modules/contrib/CMakeFiles/opencv_contrib.dir/build.make:374: recipe for target 'modules/contrib/CMakeFiles/opencv_contrib.dir/src/rgbdodometry.cpp.o' failed
make[2]: *** [modules/contrib/CMakeFiles/opencv_contrib.dir/src/rgbdodometry.cpp.o] Error 1
CMakeFiles/Makefile2:5942: recipe for target 'modules/contrib/CMakeFiles/opencv_contrib.dir/all' failed
make[1]: *** [modules/contrib/CMakeFiles/opencv_contrib.dir/all] Error 2
Makefile:160: recipe for target 'all' failed
make: *** [all] Error 2

编译顺利多了，但问题对比了一下，还是没变。
fatal error: unsupported/Eigen/MatrixFunctions: No such file or directory
compilation terminated.

解决方案是将对应的文件夹复制到/usr/include/下，举例Eigen安装到了/usr/include/eigen3/，则执行：

$ cd /usr/include/eigen3/
$ sudo cp -r unsupported/ ..
$ sudo cp -r Eigen/ ..

之后重新make OpenCV,结果显示：

[ 96%] Linking CXX executable ../../bin/opencv_perf_superres
[ 96%] Built target opencv_perf_superres
[ 97%] Built target opencv_videostab_pch_dephelp
[ 98%] Built target pch_Generate_opencv_videostab
[ 98%] Linking CXX shared library ../../lib/libopencv_videostab.so
[ 98%] Built target opencv_videostab
[ 99%] Built target opencv_haartraining_engine
[ 99%] Linking CXX executable ../../bin/opencv_performance
[ 99%] Built target opencv_performance
[ 99%] Linking CXX executable ../../bin/opencv_createsamples
[ 99%] Built target opencv_createsamples
[ 99%] Linking CXX executable ../../bin/opencv_haartraining
[ 99%] Built target opencv_haartraining
[ 99%] Linking CXX executable ../../bin/opencv_traincascade
[100%] Built target opencv_traincascade
[100%] Linking CXX executable ../../bin/opencv_annotation
[100%] Built target opencv_annotation
[100%] Linking CXX executable ../../bin/opencv_visualisation
[100%] Built target opencv_visualisation

看来这种方法是有效的！