Ubuntu18.04配置Cube-SLAM

Code：https://github.com/shichaoy/cube_slam

Notes：本文描述笔者再Ubuntu18.04安装Cube-SLAM遇到的相关问题和解决方法。

配置相关问题或者Object SLAM相关科研问题欢迎通过邮件联系笔者交流：lidong8421bcd@gmail.com（基本保证两个工作日内回复邮件）

配置过程

1.环境配置

ROS环境配置：

mkdir -p ~/catkin_ws_Cube-SLAM/src
cd catkin_ws_Cube-SLAM/src/
catkin_init_workspace
cd ..
catkin_make
source devel/setup.sh
catkin_make install
echo "source ~/catkin_ws_Cube-SLAM/devel/setup.bash" >> ~/.bashrc
1
2
3
4
5
6
7
8

代码下载：

cd src/
git clone https://github.com/shichaoy/cube_slam.git
1
2

2.代码编译

编译依赖项：

cd cube-slam/
sh install_dependenices.sh
1
2

编译工作空间：

 cd ../..
 catkin_make -j4
1
2

catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/src/ 下面文件报错如下：

error: ‘std::vector’ has not been declared
error: expected ‘,’ or ‘...’ before ‘<’ token
error: ‘all_strings’ was not declared in this scope
1
2
3

分析：看github的issue，这个问题是Ubuntu18.04编译经常会出现的问题。

解决方法：

• 1 ：在功能包的CMAKELIST中添加 set(CMAKE_CXX_STANDARD 14)，在几个C++文件中添加：#include ， error: ‘iota’ is not a member of ‘std’ 报错问题没了

• 2：在以下几个文件中添加：#include 相关报错问题解决

catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/src/box_proposal_detail.cpp
catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/src/object_3d_util.cpp
catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/src/matrix_utils.cpp
catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/include/detect_3d_cuboid/detect_3d_cuboid.h
catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/include/detect_3d_cuboid/matrix_utils.h
catkin_ws_Cube-SLAM/src/cube_slam/detect_3d_cuboid/include/detect_3d_cuboid/object_3d_util.h
1
2
3
4
5
6

• 3 ：orb_object_slam下C++文件报错：error: ‘usleep’ was not declared in this scope，经典ORB-SLAM2报错问题

解决方法：在相关报错的文件上加入头文件 #include。

#以下文件要添加#include
catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/src/LocalMapping.cc
catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/src/LoopClosing.cc
catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/src/System.cc
catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/src/Viewer.cc
1
2
3
4
5

完成之后编译成功。

运行：roslaunch object_slam object_slam_example.launch

demo可以正常跑。

3.数据集下载

这里因为作者使用的数据集有格式要求，暂且使用作者给的两个数据集（作者处理好了相关格式），后续会自己制作数据集测试。 这里作者给了两个数据集，分别对应静态情况和动态情况。

（1）Static datasets

• 在https://drive.google.com/open?id=1FrBdmYxrrM6XeBe_vIXCuBTfZeCMgApL链接上下载，数据集名字为：Seq_07

（2）Dynamic datasets

• 在 https://drive.google.com/drive/folders/1T2PmK3Xt5Bq9Z7UhV8FythvramqhOo0a?usp=sharing下载，数据集名字为：0018

这里我将两个数据集解压之后放到该路径下：catkin_ws_Cube-SLAM/src/cube_slam/Datasets

4.设置路径启动代码

（1）跑Static datasets

• 这里需要先将mono.launch文件中 中的路径修改为自己数据集的路径，我这里修改为：/home/cvpr/catkin_ws_Cube-SLAM/src/cube_slam/Datasets/seq_07-20220720T061819Z-001/seq_07

• 然后报错如下：

Loading ORB Vocabulary. This could take a while...
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
[ros_mono-3] process has died [pid 1257, exit code -6, cmd /home/cvpr/catkin_ws_Cube-SLAM/devel/lib/orb_object_slam/ros_mono /home/cvpr/catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/Vocabulary/ORBvoc.bin /home/cvpr/catkin_ws_Cube-SLAM/src/cube_slam/orb_object_slam/Examples/Monocular/KITTI04-12_me.yaml /camera/image_raw:=/kitti/left/image_raw __name:=ros_mono __log:=/home/cvpr/.ros/log/48a4443c-07f6-11ed-a895-74ee2aeddddc/ros_mono-3.log].
log file: /home/cvpr/.ros/log/48a4443c-07f6-11ed-a895-74ee2aeddddc/ros_mono-3*.log
1
2
3
4
5

这里是词袋文件的问题，即mono.launch中 args="$(find orb_object_slam)/Vocabulary/ORBvoc.bin 并没有找到这个文件；

解决方法：这里 cube_slam/orb_object_slam/Vocabulary 提供了将txt词袋转换为二进制的bin词袋的代码。这里我使用这个代码进行词袋转换有点问题，索性就直接下载了一个二进制词袋：https://github.com/nxyzgf/sgg-dbow2/blob/master/ORBvoc.bin

• 运行时，仍然报错：Gtk-ERROR **: 04:42:33.691: GTK+ 2.x symbols detected. Using GTK+ 2.x and GTK+ 3 in the same process is not supported；这里这应该是linux系统同时安装了GTK+ 2.x 和 GTK+ 3导致的，参考:https://blog.csdn.net/weixin_39768247/article/details/111855987/，卸载opencv之后又重新装一边，同时将ROS的workspace中的build和devel删除重新编译。

运行：roslaunch orb_object_slam mono.launch 正常。

然后跑ROS的bag数据集：rosbag play left_full_gray.bag

（2）跑Dynamic datasets

• 先修改mono_dynamic.launch文件：将文件中 改为自己数据集的地址：/home/cvpr/catkin_ws_Cube-SLAM/src/cube_slam/Datasets/0018-20220720T062223Z-001/0018 即可

运行：

roslaunch orb_object_slam mono_dynamic.launch 
rosbag play mono.bag 
1
2

效果如下：