文章目录

0 引言
1 安装依赖
- 1.1 opencv安装
- 1.2 Eigen3安装
- 1.3 Pangolin安装
- 1.4 其他
2 编译安装ORB-SLAM2
- 2.1 build.sh
- 2.2 build_ros.sh

0 引言

ORB-SLAM2是一种用于单目、双目和RGB-D相机的视觉SLAM（同时定位与地图构建）系统。它由萨拉戈萨大学和伦敦帝国学院的研究人员开发，并于2016年首次发布。

该系统基于ORB特征检测器和描述子，这是一种快速且具有鲁棒性的特征提取方法。ORB-SLAM2使用了多种技术，包括相机姿态估计、特征匹配、回环检测和地图优化等，以实现高精度的定位和地图构建。它还支持实时运行，并且可以在没有先验地图的情况下进行自适应构建。ORB-SLAM2已成为许多研究和工业应用中的常用SLAM系统之一。

本文主要包括Ubuntu20.04系统下，并已安装对应的ROS-noetic机器人系统，在此基础上如何一步步安装ORB-SLAM2的依赖库，比如OpenCV，Eigen3，Pangolin，最后安装非ROS版本和ROS版的ORB-SLAM2，以及安装过程中遇到的问题和解决问题的办法。

本文系统：

Ubuntu20.04系统
ROS-noetic机器人系统

1 安装依赖

1.1 opencv安装

建议下载源码编译安装的方式，OpenCV-release下载地址，可以选择OpenCV3的版本，比如OpenCV-3.4.16版本，找到对应版本，点击Sources下载即可。

# 安装OpenCV依赖
sudo apt-get install build-essential libgtk2.0-dev libgtk-3-dev libavcodec-dev libavformat-dev libjpeg-dev libswscale-dev libtiff5-dev

# 进入到OpenCV下载文件夹目录，新开终端，这里以3.4.16版本为例
unzip opencv-3.4.16.zip
# 编译安装
cd opencv-3.4.16/
mkdir build
cd build
# cmake
cmake -D CMAKE_BUILD_TYPE=Release -D OPENCV_GENERATE_PKGCONFIG=YES ..
# make
make -j4
# install
sudo make install

安装后可进行测试

cd opencv-3.4.16/samples/cpp/example_cmake
cmake .
make
./opencv_example
# 笔记本有摄像头的话，会自动打开，如下图

请添加图片描述
也可以用以下命令查看安装的版本

pkg-config --modversion opencv
# 3.4.16

1.2 Eigen3安装

eigen3 一般已经安装了，可用如下命令查看：

# 查看前18行，可自行调整
cat /usr/include/eigen3/Eigen/src/Core/util/Macros.h | head -n 18

比如出现如下所示，eigen版本是3.3.7

#ifndef EIGEN_MACROS_H
#define EIGEN_MACROS_H#define EIGEN_WORLD_VERSION 3
#define EIGEN_MAJOR_VERSION 3
#define EIGEN_MINOR_VERSION 7

如果确定没有安装，可用如下命令直接安装：

sudo apt-get install libeigen3-dev

1.3 Pangolin安装

同样也是源码编译安装，Pangolin下载地址，本文用的是Pangolin-0.6版本，进入后选择v0.6，然后点击source code(zip)下载即可。

# 安装依赖库
sudo apt-get update
sudo apt-get install libxkbcommon-dev
sudo apt-get install wayland-protocols
sudo apt install libglew-dev

# 进入到Pangolin下载文件夹目录，新开终端，这里以0.6版本为例
unzip Pangolin-0.6.zip
cd Pangolin-0.6/
mkdir build
cd build
# cmake
cmake ..
# make
make -j4
# install
sudo make install

安装可能出现的问题：

make时缺少pybind11.h的报错问题

fatal error: pybind11/pybind11.h: No such file or directory30 | #include <pybind11/pybind11.h>

解决办法：由于下载的Pangolin-0.6.zip，解压后external/pybind11文件夹下是空的，所以找不到，当然CMakeLists.txt中也有find_package(pybind11 QUIET)，所以有以下两种办法解决，本文用第二种解决了编译错误。

1.安装pybind11：python3 -m pip install pybind11 或 pip install pybind11
2.下载pybind11源码解压后放到external/pybind11文件夹下，比如下载了pybind11-2.10.4.zip，解压的pybind11-2.10.4文件夹中的所有文件放到external/pybind11文件夹下

1.4 其他

DBoW2 and g2o两个库已经包含在官方工程的ORB-SALM2-->Thirdparty文件夹中，同时官方也把这两个库的安装命令写到了编译安装脚本build.sh中。
ORB-SLAM2也有ROS版本，所以也需要安装ROS机器人系统，如果是Ubuntu20.04，那么就安装对应的ROS-noetic版本。

2 编译安装ORB-SLAM2

2.1 build.sh

以下是非ROS版本的ORB_SLAM2安装：

# 比如home下新开终端，确定下载的路径
# 下载ORB_SLAM2
git clone https://github.com/raulmur/ORB_SLAM2.git
cd ORB_SLAM2
chmod +x build.sh
./build.sh

编译出现的问题：
问题1：error: ‘usleep’ was not declared in this scope

解决办法：ORB_SLAM2/include/System.h头文件中添加

#include<unistd.h>

问题2：LoopClosing.h 中eigen类别定义问题

解决办法：ORB_SLAM2/include/ORBextractor.h中如下修改

// Eigen::aligned_allocator<std::pair<const KeyFrame*, g2o::Sim3> > > KeyFrameAndPose;
Eigen::aligned_allocator<std::pair<KeyFrame* const, g2o::Sim3> > > KeyFrameAndPose;

问题3：OpenCV头文件引用问题
解决办法：ORB_SLAM2/include/ORBextractor.h中注释掉cv.h，新增如下两行

// #include <opencv/cv.h>
#include <opencv2/imgproc/imgproc_c.h>
#include <opencv2/highgui/highgui_c.h>

2.2 build_ros.sh

以下是ROS版的ORB_SLAM2安装：
打开build_ros.sh文件中可以看到，ROS版在ORB_SLAM2/Examples/ROS/ORB_SLAM2/目录下，对应的编译build也是在对应的文件夹下。

# 冒号后的 PATH 替换成自己的ORB_SLAM2工程保存目录
# 添加以下export到 ~/.bashrc 中，可以 vim ~/.bashrc,添加到最后一行
export ROS_PACKAGE_PATH=${ROS_PACKAGE_PATH}:PATH/ORB_SLAM2/Examples/ROS
source ~/.bashrc
# 给.sh脚本权限
chmod +x build_ros.sh
# 编译
./build_ros.sh

编译可能出现的问题：
问题1：ROS-noetic中OpenCV4和ORB_SLAM2中OpenCV3不匹配的问题

/usr/bin/ld: warning: libopencv_imgcodecs.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_imgproc.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_imgproc.so.3.4
/usr/bin/ld: warning: libopencv_core.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_core.so.3.4
/usr/bin/ld: CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: undefined reference to symbol '_ZNK2cv8FileNodecviEv'
/usr/bin/ld: /usr/lib/x86_64-linux-gnu/libopencv_core.so.4.2.0: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
make[2]: *** [CMakeFiles/Stereo.dir/build.make:179: ../Stereo] Error 1
make[1]: *** [CMakeFiles/Makefile2:514: CMakeFiles/Stereo.dir/all] Error 2
make: *** [Makefile:130: all] Error 2

解决办法：
因为ORB_SLAM2用的OpenCV3，刚刚1.1小节中安装的也是OpenCV3，但是Ubuntu20.04对应的ROS noetic自带的是OpenCV4，而cv_bridge链接的是OpenCV4，所以，与OpenCV3存在冲突，解决方法如下:

根据报错提示，进入libopencv_imgcodecs.so.4.2和libopencv_proc.so.4.2所在位置/usr/lib/x86_64-linux-gnu/目录下，然后复制到ORB_SLAM2的ROS版本的lib目录中

# PATH替换成自己的ORB_SLAM2工程保存目录
cd /usr/lib/x86_64-linux-gnu/
sudo cp libopencv_imgproc.so.4.2 PATH/ORB_SLAM2/lib/libopencv_imgproc.so.4.2
sudo cp libopencv_core.so.4.2 PATH/ORB_SLAM2/lib/libopencv_core.so.4.2

然后在ORB_SLAM2/Examples/ROS/ORB_SLAM2/CMakeLists.txt中添加两个lib的路径${PROJECT_SOURCE_DIR}/../../../lib/libopencv_core.so.4.2和${PROJECT_SOURCE_DIR}/../../../lib/libopencv_imgproc.so.4.2

# ORB_SLAM2/Examples/ROS/ORB_SLAM2/CMakeLists.txt文件中添加
set(LIBS 
${OpenCV_LIBS} 
${EIGEN3_LIBS}
${Pangolin_LIBRARIES}
${PROJECT_SOURCE_DIR}/../../../Thirdparty/DBoW2/lib/libDBoW2.so
${PROJECT_SOURCE_DIR}/../../../Thirdparty/g2o/lib/libg2o.so
${PROJECT_SOURCE_DIR}/../../../lib/libORB_SLAM2.so
# for opencv4
${PROJECT_SOURCE_DIR}/../../../lib/libopencv_core.so.4.2
${PROJECT_SOURCE_DIR}/../../../lib/libopencv_imgproc.so.4.2

然后回到ORB_SLAM2的工程目录下，重新执行./build_ros.sh，如下，不再报错，编译成功。

[ 22%] Linking CXX executable ../MonoAR
[ 22%] Linking CXX executable ../RGBD
[ 33%] Linking CXX executable ../Mono
[ 44%] Linking CXX executable ../Stereo
/usr/bin/ld: warning: libopencv_imgcodecs.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_core.so.3.4, needed by ../../../../lib/libORB_SLAM2.so, may conflict with libopencv_core.so.4.2
/usr/bin/ld: warning: libopencv_imgcodecs.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_core.so.3.4, needed by ../../../../lib/libORB_SLAM2.so, may conflict with libopencv_core.so.4.2
/usr/bin/ld: warning: libopencv_core.so.3.4, needed by /usr/local/lib/libopencv_highgui.so.3.4.16, may conflict with libopencv_core.so.4.2
/usr/bin/ld: warning: libopencv_imgcodecs.so.4.2, needed by /opt/ros/noetic/lib/libcv_bridge.so, may conflict with libopencv_imgcodecs.so.3.4
/usr/bin/ld: warning: libopencv_core.so.3.4, needed by ../../../../lib/libORB_SLAM2.so, may conflict with libopencv_core.so.4.2
[100%] Built target Stereo
[100%] Built target RGBD
[100%] Built target Mono
[100%] Built target MonoAR

问题2：rosdep的问题

[rospack] Error: could not call python function 'rosdep2.rospack.init_rospack_interface'CMake Error at /opt/ros/melodic/share/ros/core/rosbuild/public.cmake:129 (message):Failed to invoke rospack to get compile flags for package 'ORB_SLAM2'.Look above for errors from rospack itself.  Aborting.  Please fix thebroken dependency!Call Stack (most recent call first):/opt/ros/melodic/share/ros/core/rosbuild/public.cmake:207 (rosbuild_invoke_rospack)CMakeLists.txt:4 (rosbuild_init)-- Configuring incomplete, errors occurred!

解决办法：
手动执行修复rosdep

sudo rosdep fix-permissions

但解决这个问题后，又出现下述的问题3…

问题3：libboost_system.so链接不到的问题

collect2: error: ld returned 1 exit status
/usr/bin/ld: CMakeFiles/Stereo.dir/src/ros_stereo.cc.o: undefined reference to symbol '_ZN5boost6system15system_categoryEv'
/usr/lib/x86_64-linux-gnu/libboost_system.so: error adding symbols: DSO missing from command line

解决办法：
手动添加-lboost_system到ORB_SLAM2/Examples/ROS/ORB_SLAM2/CMakeLists.txt，如下所示：

set(LIBS 
${OpenCV_LIBS} 
${EIGEN3_LIBS}
${Pangolin_LIBRARIES}
${PROJECT_SOURCE_DIR}/../../../Thirdparty/DBoW2/lib/libDBoW2.so
${PROJECT_SOURCE_DIR}/../../../Thirdparty/g2o/lib/libg2o.so
${PROJECT_SOURCE_DIR}/../../../lib/libORB_SLAM2.so
# 解决libboost_system.so链接不到的问题
-lboost_system

然后回到ORB_SLAM2的工程目录下，重新执行./build_ros.sh，如下，不再报错，编译成功。

[ 11%] Linking CXX executable ../Mono
[ 22%] Linking CXX executable ../RGBD
[ 33%] Linking CXX executable ../MonoAR
[ 44%] Linking CXX executable ../Stereo
[ 55%] Built target RGBD
[ 66%] Built target Mono
[ 88%] Built target MonoAR
[100%] Built target Stereo

Reference:

https://github.com/raulmur/ORB_SLAM2
https://stackoverflow.com/questions/46396204/affdex-cpp931-fatal-error-pybind11-pybind11-h-no-such-file-or-directory-in
https://stackoverflow.com/questions/61758906/building-cpp-with-pybind11

须知少时凌云志，曾许人间第一流。

⭐️👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍🌔