Skip to content

Latest commit

 

History

History
350 lines (188 loc) · 4.58 KB

README.md

File metadata and controls

350 lines (188 loc) · 4.58 KB
Raw

Custom Datasets

Generate Dataset with ZED camera

ZED启动

方式一：使用ZED SDK

  • 设置分辨率**：修改zed_ws/src/zed-ros-wrapper/zed_wrapper/params/common.yaml中的general中的resolution参数， **0: HD2K, 1: HD1080, 2: HD720, 3: VGA

  • launch ZED camera

roslaunch zed_wrapper zed.launch
  • 查看图像话题
rostopic list
  • 查看话题的分辨率
rostopic echo /zed/zed_node/stereo_raw/image_raw_color --noarr

方式二：当做UVC相机使用

rosrun pub_zed pub_zed

记录传感器信息

录制为ROS bag

这里录制的话题包括：

/imu/data
/imu/mag
/odom
/odometry/filtered
/scan
/zed_cam/cam0
/zed_cam/cam1

执行：

rosbag record /imu/data \
/imu/mag \
/odom \
/odometry/filtered \
/scan \
/zed_cam/cam0 \
/zed_cam/cam1 \
-o room.bag
  • rviz查看录制成功后的bag
#将bag从机器人移动到本地
scp -r [email protected]:~/room_2022-12-08-23-05-00.bag   /home/chen/datasets/MyData/bags

rosbag play room_2022-12-7.bag

rviz -d  ${dynamic_vins_root}/src/custom_dataset/config/rviz/custom_dataset.rviz

记录VICON轨迹

  • 启动Vicon的接口
roslaunch vicon_bridge vicon.launch
  • 记录轨迹
rosrun sub_vicon sub_vicon "room"
  • 可视化vicon轨迹
startconda && conda activate py36

evo_traj tum vicon.txt -p

记录Gmapping轨迹

建图

  • 运行gampping
#在机器人上运行
roslaunch ir100_description load.launch

roslaunch ir100_navigation gmapping_demo.launch
  • 运行rviz
#在本地主机上运行
rviz -d /home/chen/ws/robot_exp_ws/src/gmapping_demo/rviz/gmapping.rviz
  • 保存地图
rosrun map_server map_saver -f map_426b

导航

  • 启动导航，加载之前保存的地图
roslaunch ir100_navigation ir100_navigation.launch
  • 开始保存机器人的轨迹
rosrun tf_test write_tf_stamp \odom \base_link  gmapping.txt
  • 查看机器人保存的轨迹
startconda && conda activate py36
evo_traj tum gmapping.txt -p

或
evo_traj tum --align  -p gmapping.txt --ref vicon.txt -s

轨迹评估

  • 评估VICON和Gmapping的轨迹
pose_path=gmapping.txt
vicon_path=vicon.txt

evo_ape tum --align  ${pose_path} ${vicon_path} && evo_rpe tum --align  ${pose_path} ${vicon_path} -r trans_part && evo_rpe tum --align  ${pose_path} ${vicon_path} -r rot_part

录制步骤总结

  • 设备启动：机器人、本地PC、Vicon，且均连接到机器人的局域网下

  • [PC]启动Vicon的接口

roslaunch vicon_bridge vicon.launch
  • [R]启动GMapping定位
  • [R]启动ZED相机
rosrun pub_zed pub_zed
  • [PC]同时录制机器人轨迹和vicon的轨迹
#roslaunch sub_vicon record.launch

rosrun sub_vicon sub_vicon "room" true

rosrun tf_test write_tf_stamp \odom \base_link  gmapping.txt
  • [R]录制传感器信息
rosbag record /imu/data \
/imu/mag \
/odom \
/odometry/filtered \
/scan \
/zed_cam/cam0 \
/zed_cam/cam1 \
-o room.bag

上面的PC表示本地主机，R表示机器人。

传感器标定

自定义

Offline process Zed dataset

Convert bag to images

将图像话题保存为图像文件

  • 修改custom_dataset中的sub_write_images.cpp

  • 编译custom_dataset

  • 记录bag中左相机的图像

source  ${dynamic_vins_root}/devel/setup.bash  

rosrun custom_dataset sub_write_images /home/chen/datasets/MyData/ZED_data/corridor_dynamic_1/cam0
  • 播放数据集
rosbag play -r 0.5 ./room_dynamic_1/data.bag

Instance Segmentation

  • 进行实例分割
python solov2_det2d_zed.py

Generate Dataset with MYNT-Eye-Camera

启动相机

  • 首先安装MYNT-eye的SDK
MYNT_EYE_PATH=~/app/MYNT-EYE-S-SDK

source ${MYNT_EYE_PATH}/wrappers/ros/devel/setup.bash

roslaunch mynt_eye_ros_wrapper vins_fusion.launch

查看消息

查看消息频率

rostopic hz /mynteye/left_rect/image_rect

查看消息内容

rostopic echo /mynteye/left_rect/image_rect --noarr

录制

rosbag record \
/mynteye/disparity/camera_info \
/mynteye/disparity/image_raw \
/mynteye/left_rect/camera_info \
/mynteye/left_rect/image_rect \
/mynteye/right_rect/camera_info \
/mynteye/right_rect/image_rect \
/mynteye/imu/data_raw \
-o room.bag  --duration=300
  • 复制到本地主机
scp -r ${robotip}:~/room_2023-01-07-15-27-02.bag   /home/chen/datasets/MyData/bags/