add bag files for color space detection

[k-okada]

@tongtybj, @chibi314
please upload your bag files for color space detection which you used in ros-perception/opencv_apps#61 (comment)
If you have several bag files with different lighting condition, that's fine.
If you can provide bag file taken by real drone, with IMU data, that's much better. If not, bagfile with only images is fine for now.

@wkentaro will help you how to upload bag data

[wkentaro]

Please refer to

• [jsk_pcl_ros] ICP Registration on 2D plane #1991 (comment)
• http://jsk-docs.readthedocs.io/en/latest/jsk_common/doc/jsk_data/scripts/jsk_data_cli.html

[tongtybj]

@wkentaro

I got following error when trying to upload:

$ jsk_data put --public 2017-04-14_uav_with_marked_tree_kashiwa_camp_strong_light_industrial_cam.bag.tgz
Uploading to Google Drive...
Go to the following link in your browser:
https://accounts.google.com/o/oauth2/auth?client_id=367116221053-7n0vf5akeru7on6o2fjinrecpdoe99eg.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&response_type=code&scope=https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive&state=state

Enter verification code: An error occurred creating Drive client: OAuthError: updateToken: Unexpected HTTP status 400 Bad Request

An error occurred creating Drive client: OAuthError: updateToken: Unexpected HTTP status 400 Bad Request
Traceback (most recent call last):
  File "/opt/ros/kinetic/bin/jsk_data", line 8, in <module>
    jsk_data.cli.cli(obj={})
  File "/usr/lib/python2.7/dist-packages/click/core.py", line 716, in __call__
    return self.main(*args, **kwargs)
  File "/usr/lib/python2.7/dist-packages/click/core.py", line 696, in main
    rv = self.invoke(ctx)
  File "/usr/lib/python2.7/dist-packages/click/core.py", line 1060, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/usr/lib/python2.7/dist-packages/click/core.py", line 889, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/usr/lib/python2.7/dist-packages/click/core.py", line 534, in invoke
    return callback(*args, **kwargs)
  File "/opt/ros/kinetic/lib/python2.7/dist-packages/jsk_data/cli.py", line 142, in cmd_put
    stdout = upload_gdrive(filename)
  File "/opt/ros/kinetic/lib/python2.7/dist-packages/jsk_data/gdrive.py", line 72, in upload_gdrive
    return run_gdrive(args=args)
  File "/opt/ros/kinetic/lib/python2.7/dist-packages/jsk_data/gdrive.py", line 30, in run_gdrive
    return subprocess.check_output(cmd, shell=True)
  File "/usr/lib/python2.7/subprocess.py", line 574, in check_output
    raise CalledProcessError(retcode, cmd, output=output)
subprocess.CalledProcessError: Command 'rosrun jsk_data drive-linux-x64 --config /home/chou/.ros/jsk_data/.gdrive upload --file 2017-04-14_uav_with_marked_tree_kashiwa_camp_strong_light_industrial_cam.bag.tgz --parent 0B9P1L--7Wd2vUGplQkVLTFBWcFE' returned non-zero exit status 1

[wkentaro]

@tongtybj Have you done this?

Go to the following link in your browser:
https://accounts.google.com/o/oauth2/auth?client_id=367116221053-7n0vf5akeru7on6o2fjinrecpdoe99eg.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&response_type=code&scope=https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive&state=state

[tongtybj]

@wkentaro
Thx! WFM

[tongtybj]

@k-okada
I upload 5 rosbags with different scenes and image sensors.

Image Sensor: pointgrey chameleon3, web camera.

Scene1: hongo sanshiro forest with pointgrey chameleon3
rosbag: 2017-04-16_uav_with_marked_tree_hongo_sanshiro_forest_light_industrial_cam.bag

Scene2: kashiwa camp(ground) with pointgrey chameleon3
rosbag: 2017-04-14_uav_with_marked_tree_kashiwa_camp_strong_light_industrial_cam.bag.tgz

Scene3: kashiwa camp(ground) with web camera
rosbag: 2017-04-19_uav_with_marked_tree_kashiwa_camp_strong_light_web_cam.bag

Scene4: kashiwa camp(forest) with web camera
rosbag: 2017-04-21_uav_with_marked_tree_kashiwa_camp_forest_light_web_cam1.bag
rosbag: 2017-04-21_uav_with_marked_tree_kashiwa_camp_forest_light_web_cam2.bag

Scene5: indoor(Eng.8 R.330) with web camera
rosbag: 2017-04-21_uav_with_marked_tree_inside_web_cam.bag

[tongtybj]

@k-okada
All rosbag files contains the imu and odometry data of UAV, along with the range sensor(ultrasonic) to the ground.

$ rosbag info 2017-04-21_uav_with_marked_tree_kashiwa_camp_forest_light_web_cam2.bag
path:        2017-04-21_uav_with_marked_tree_kashiwa_camp_forest_light_web_cam2.bag
version:     2.0
duration:    2:02s (122s)
start:       Apr 21 2017 14:28:15.17 (1492752495.17)
end:         Apr 21 2017 14:30:17.39 (1492752617.39)
size:        2.7 GB
messages:    99284
compression: none [3105/3105 chunks]
types:       dji_sdk/Acceleration         [5de5cfee671950d30a03b944f0d1555c]
             dji_sdk/AttitudeQuaternion   [999d24c7cb273aa4f2c6044f85cac84c]
             dji_sdk/GlobalPosition       [10784f0f63ab6f41e201fee714fabb2a]
             dji_sdk/LocalPosition        [933ce5db06b6bff36785c58a964ad3c7]
             geometry_msgs/PointStamped   [c63aecb41bfdfd6b7e1fac37c7cbe7bf]
             geometry_msgs/Twist          [9f195f881246fdfa2798d1d3eebca84a]
             geometry_msgs/Vector3Stamped [7b324c7325e683bf02a9b14b01090ec7]
             nav_msgs/Odometry            [cd5e73d190d741a2f92e81eda573aca7]
             sensor_msgs/CameraInfo       [c9a58c1b0b154e0e6da7578cb991d214]
             sensor_msgs/Image            [060021388200f6f0f447d0fcd9c64743]
             sensor_msgs/LaserScan        [90c7ef2dc6895d81024acba2ac42f369]
topics:      /camera/camera_info             3104 msgs    : sensor_msgs/CameraInfo      
             /camera/image_rect              3104 msgs    : sensor_msgs/Image           
             /cmd_vel                        2348 msgs    : geometry_msgs/Twist         
             /dji_sdk/acceleration          12219 msgs    : dji_sdk/Acceleration        
             /dji_sdk/attitude_quaternion   12219 msgs    : dji_sdk/AttitudeQuaternion  
             /dji_sdk/global_position       12219 msgs    : dji_sdk/GlobalPosition      
             /dji_sdk/local_position        12219 msgs    : dji_sdk/LocalPosition       
             /dji_sdk/odometry              12219 msgs    : nav_msgs/Odometry           
             /guidance/ultrasonic            2442 msgs    : sensor_msgs/LaserScan       
             /guidance/velocity              1221 msgs    : geometry_msgs/Vector3Stamped
             /modified_odom                 12210 msgs    : nav_msgs/Odometry           
             /object_image_center            3104 msgs    : geometry_msgs/PointStamped  
             /scan                           4874 msgs    : sensor_msgs/LaserScan       
             /tree_location                  3434 msgs    : geometry_msgs/PointStamped  
             /uav_target_pos                 2348 msgs    : nav_msgs/Odometry

Also note that the odometry is obtained from IMU + downwards optical flow + GPS, and recently the GPS is really suck, because US army add undecryptable noise on the GPS data(NEWS). So the accuracy of position is very bad. What we done is to simply integrate the linear velocity from optical flow for the real experiments.

I am trying the 2D laser slam, and looking forwards to test imu + monocular mapping like LSD-SLAM

[k-okada]

Nice, could you create smaller version? may be only image and color info, or (+ dji_sdk... + guidance)

$ rosbag info 2017-04-16_uav_with_marked_tree_hongo_sanshiro_forest_light_industrial_cam.bag
path:        2017-04-16_uav_with_marked_tree_hongo_sanshiro_forest_light_industrial_cam.bag
version:     2.0
duration:    48.0s
start:       Apr 16 2017 14:24:57.59 (1492320297.59)
end:         Apr 16 2017 14:25:45.54 (1492320345.54)
size:        5.3 GB
messages:    33569
compression: none [1441/1441 chunks]
types:       dji_sdk/Acceleration         [5de5cfee671950d30a03b944f0d1555c]
             dji_sdk/AttitudeQuaternion   [999d24c7cb273aa4f2c6044f85cac84c]
             dji_sdk/GlobalPosition       [10784f0f63ab6f41e201fee714fabb2a]
             dji_sdk/LocalPosition        [933ce5db06b6bff36785c58a964ad3c7]
             geometry_msgs/PointStamped   [c63aecb41bfdfd6b7e1fac37c7cbe7bf]
             geometry_msgs/Twist          [9f195f881246fdfa2798d1d3eebca84a]
             geometry_msgs/Vector3Stamped [7b324c7325e683bf02a9b14b01090ec7]
             nav_msgs/Odometry            [cd5e73d190d741a2f92e81eda573aca7]
             sensor_msgs/CameraInfo       [c9a58c1b0b154e0e6da7578cb991d214]
             sensor_msgs/Image            [060021388200f6f0f447d0fcd9c64743]
             sensor_msgs/LaserScan        [90c7ef2dc6895d81024acba2ac42f369]
topics:      /camera/camera_info            1439 msgs    : sensor_msgs/CameraInfo      
             /camera/image_rect             1440 msgs    : sensor_msgs/Image           
             /cmd_vel                        959 msgs    : geometry_msgs/Twist         
             /dji_sdk/acceleration          4796 msgs    : dji_sdk/Acceleration        
             /dji_sdk/attitude_quaternion   4796 msgs    : dji_sdk/AttitudeQuaternion  
             /dji_sdk/global_position       4796 msgs    : dji_sdk/GlobalPosition      
             /dji_sdk/local_position        4796 msgs    : dji_sdk/LocalPosition       
             /dji_sdk/odometry              4795 msgs    : nav_msgs/Odometry           
             /guidance/ultrasonic            959 msgs    : sensor_msgs/LaserScan       
             /guidance/velocity              479 msgs    : geometry_msgs/Vector3Stamped
             /object_image_center           1440 msgs    : geometry_msgs/PointStamped  
             /scan                          1915 msgs    : sensor_msgs/LaserScan       
             /uav_target_pos                 959 msgs    : nav_msgs/Odometry

odometry is obtained from IMU + downwards optical flow + GPS
imu + monocular mapping like LSD-SLAM

When you integrate multiple sensor data, you need check accuracy and characteristics in step-by-step manner. And before you merge data, you have to image how each data compensate each other.
For example, usually optical flow is not s accurate at least in-house environment, it may work well on outdoor scene, but it may depends on camera motion speeds vs search area in flow.
So, first you should check only with IMU as http://wiki.ros.org/robot_pose_ekf#How_Robot_Pose_EKF_works, and then you consider what kind of information is missing and choose appropriate sensor which covers that information.

[tongtybj]

@k-okada
Thank you for your advice!

I will check step by step.
As for the optical flow sensor, it is a mature product form DJI, called guidance, which has their own fusion system inside the black box using IMU + Stereo Optical Flow + Ultra-sonic.