Veerachart's profile - activity

Tracking while the person is rotating seems somehow working well. Anyway, it always thinks that you enter facing away from the camera, even you are looking at the camera

I also found that the joints aren't rotating, they only show positions. The first version gives some clues about the direction.

@Chaos I finished installing Nite2 and openni2_tracker package, after some edits of the code. Now I can run it with Asus Xtion, but not with Kinect. It says "Can't open device." and crashes.

Now the published frame seems to be mirrored with the depth image at this moment.

@Chaos the link for Nite2 there is broken ( openni.org is already closed for some time.)

I installed openni2_camera and tried it with Kinect; not working. Worked well with Asus Xtion.

I tried catkin_make your package, but it said Nite2 is not found.

Hi @Chaos, I have never tried that. I searched the wiki, and found openni2_launch package, but it says that it is for RGB-D sensors other than Kinect, which I am using. The package openni2_tracker seems to be only in hydro, but I'm using groovy. If it is usable on groovy & Kinect, I'll try tomorrow.

Hi guys,

I have been using openni_tracker for a while, using the position of some joints and direction of the person. However, now I add more rotation to the person, and the direction of the person seems messed up. After trying rotation and observe closely, I saw that when the person's side just rotates pass the Kinect, left and right joints are switched, making the direction of the person also switched.

So it is:

Before rotation

left | right

After 180 degree

left | right (should be right | left )

And after another 180 degree

left | right

I also tried rotating while lifting my right hand up, and now the left and right joints are not messed. I think it is because the lifted right hand provide something for the tracker to track.

Does anyone have any solutions about correctly detecting person's direction?

I got the same thing using Kinects on ROS Groovy (using the catkinized version). I tried getting the type of both inputs to cv.Resize() by

type(self.bridge.imgmsg_to_cv(self.image, 'rgb8') == type(window)

and it returns true.

Edit: this happens only when I use IR camera. It works with RGB.

So, tf_echo checks for that and does not display it? Is there any function to remove a frame (I mean, manually or automatically)?

Now I already got it working by using gscam package, with this launch file from the v4l example in the package, removing the framerate option:

<launch>
    <!-- This launchfile should bring up a node that broadcasts a ros image
    transport on /webcam/image_raw -->
    <arg name="DEVICE" default="/dev/video0"/>
    <!-- The GStreamer framerate needs to be an integral fraction -->
    <arg name="FPS" default="30/1"/>
    <arg name="PUBLISH_FRAME" default="false"/>
    <node ns="v4l" name="gscam_driver_v4l" pkg="gscam" type="gscam" output="screen">
        <param name="camera_name" value="default"/>
        <param name="camera_info_url" value="package://gscam/examples/uncalibrated_parameters.ini"/>
        <param name="gscam_config" value="v4l2src device=$(arg DEVICE) ! video/x-raw-rgb! ffmpegcolorspace"/>
        <param name="frame_id" value="/v4l_frame"/>
        <param name="sync_sink" value="true"/>
    </node>
    <node if="$(arg PUBLISH_FRAME)" name="v4l_transform" pkg="tf" type="static_transform_publisher" args="1 2 3 0 -3.141 0 /world /v4l_frame 10"/>
</launch>

With this, topic /v4l/camera/image_raw is available for further use.

Thanks for suggestions @Eric Perko.

@Eric Perko I tried and guvcview cannot start with the capture device attached, so it is not compatible with UVC.

If it's not compatible to UVC, is there any other driver that I can try?

That does not work, with the same error of pixel format unavailable.

v4l2-ctl --list-formats-ext

gives

ioctl: VIDIOC_ENUM_FMT
Index       : 0
Type        : Video Capture
Pixel Format: 'YUYV'
Name        : 16 bpp YUY2, 4:2:2, packed

http://www.ideasonboard.org/uvc/ not supported?

Thanks for the suggestion. I'll try this as soon as I have the chance and update the result.

Hi guys,

I'm using a USBTV007 EasyCap (ID 1b71:3002 by lsusb) to capture wireless video. The system can already see it in /dev/ as video0 and I can use VLC media player to display the video on the screen and even record the video to a file. However, when I tried using uvc_camera to access the camera, this appeared:

pixfmt 0 = 'YUYV' desc = '16 bpp YUY2, 4:2:2, packed'
terminate called after throwing an instance of 'std::runtime_error'
what():  pixel format unavailable

It cannot start, and I cannot get the video to ROS.

I also tried with uvc_cam and usb_cam, and they don't work too.

I'm running Ubuntu 12.04 with Groovy. Also the kernel is updated to 3.11 (I found here that from 3.11, this model is supported; before the update the device was not seen as a video device.)

Any solutions or suggestions are appreciated.

Thank you, Veerachart

Hi guys,

My problem is that lookupTransform function can still get the transformation and rotation of a frame that has stopped publishing already (I tried tf_echo and tf_monitor and it is not there.)

What I am doing is that I use Kinect and openni_tracker to get the skeleton, then use a frame from the skeleton to set up another frame relative to the world frame. However, when the human is not present in the Kinect view anymore, the skeleton's frames are still in the list got from listener.getFrameStrings(). Anyway, my broadcaster.sendTransform(...) is not called, as far as I have checked, so my frame is not published. (tf_monitor also shows that it is not there.)

But, another node that is waiting for this frame can still get translation and rotation from lookupTransform function, which is not what I want.

Any possible causes? Thank you.

Hi,

I tried to follow the Quick Start Guide of openni_launch. As I am using Ubuntu 13.10, I have to install ROS (hydro) from source, not by apt-get, and it is the same way for other packages.

So, I installed openni_launch and openni_camera by cloning from github into my workspace, catkin_make it, and source the bash file. Then I ran

roslaunch openni_launch openni_launch.launch

And the following red texts appeared:

[ERROR] [1393915524.708745003]: Failed to load nodelet [/camera/driver] of type [openni_camera/driver]: Could not find library corresponding to plugin openni_camera/driver. Make sure the plugin description XML file has the correct name of the library and that the library actually exists.\\[FATAL] [1393915524.710317875]: Service call failed!

...

[camera/driver-2] process has died [pid 10661, exit code 255, cmd /home/veerachart/Research/ROS/ros_catkin_ws/install_isolated/lib/nodelet/nodelet load openni_camera/driver camera_nodelet_manager --no-bond ir:=ir rgb:=rgb depth:=depth depth_registered:=depth_registered projector:=projector __name:=driver __log:=/home/veerachart/.ros/log/d7834e80-a361-11e3-9fd2-648099777478/camera-driver-2.log].\\log file: /home/veerachart/.ros/log/d7834e80-a361-11e3-9fd2-648099777478/camera-driver-2*.log

(Full output here)

When trying to use Rviz to visualize, nothing appeared. It said "No image received." image_view also showed nothing,

Anyone has had this kind of problem? Is it related to the fact that I installed ROS from source, and I created another workspace separated from the one that I used to install it?

My system: Ubuntu 13.10 64-bit, ros-hydro installed from source, Kinect for Xbox (but I have tried with Kinect for Windows too)

Thank you,

Veerachart

It's Panasonic CF-AX2TETBR, maybe only sold in Japan. By the way, the error is the same even I disconnect my Kinect before launching the launch file. I have put my full output here: http://pastebin.com/WYAhB6cq

eSATA is not written near my USB port, only SS for USB3.0 symbol.

One thing that I notice is that in the error, there are 2 "/" in the address "/home/veerachart/Research/catkin_ws/devel/lib//libopenni_nodelet.so". Would that cause the problem. Anyway I haven't changed anything from the original package, even tried removing the library and download it & make again.

I tried the "killall" command, and it said: "XnSensorServer: no process found". Changing port also didn't help. I don't understand about usb2.0/esata that you mentioned. The other 2 links seem unrelevant to my problem. One more thing, I had been using the same Kinect on this notebook on USB3.0 before installing ROS, with libfreenect library.

Hi patrickl. "lsusb" show that Kinect is connected to the computer. I'm using USB3.0 port, but I have tried switching to USB2.0 from BIOS, and the same thing happened. Only camera/driver died; the rest started, and "rostopic list" still shows a lot of camera topics. I'll try the rest of you suggestion on Monday, but anyway I don't have a USB2.0 on my computer.

In case you did not install by pip, try "sudo apt-get remove python-pyparsing".

Well, I found this http://stackoverflow.com/questions/15951748/pydot-and-graphviz-error-couldnt-import-dot-parser-loading-of-dot-files-will and tried according to the top comment. It works! (You may need to add sudo.) Hope it also works with yours. (I cannot post to answer.)

Well, I forgot to add the source command in my .bashrc. After sourcing the setup.bash file the first error becomes [ERROR] [1394169434.884048329]: Failed to load nodelet [/camera/driver] of type [openni_camera/driver]: Failed to load library /home/veerachart/Research/catkin_ws/devel/lib//libopenni_nodelet.so. Make sure that you are calling the PLUGINLIB_EXPORT_CLASS macro in the library code, and that names are consistent between this macro and your XML. Error string: Could not load library (Poco exception = /home/veerachart/Research/catkin_ws/devel/lib//libopenni_nodelet.so: undefined symbol: _ZTIN7log4cxx7helpers13ObjectPtrBaseE)