Pitscho's profile - activity

Hi

I decided to do all Camera Stuff ith die image_geometry class provided by the opencv package.

Prior I rectified my camera image with:

getOptimalNewCameraMatrix initUndistortRectifyMap remap

This worked well, i could choose to show the black pixels at the edges or to cut an size the image so it only has valid pixels.

With image_geometry and the pinhole camera model class i don't see a way so rectify the image AND remove all regions with unvalid pixels.

Does someone know how?

THX

Hi

I have a subscribed topic with the type sensor_msgs::PointCloud2.

I do not need the pcl library. I just want to iterate through each point and take the Z value. but i have no clue how to do this with sensor_msgs::PointCloud2 and i dont find an example.

Do I really need pcl for easy element access? Ciao

Hi

Since the last update the topic

/camera/depth_registered/image

is gone. I used this, because the depth values were in meter.

When I disable registration i have the topic

/camera/depth/image

But this doesn't help me, because i need the registered image. It seems like the update somehow removed the /camera/depth_registered/image topic. Only /camera/depth_registered/image_raw is published.

Hi

You can set the length of the arrows in rviz.

Go to the odometry tree on the left and change the length value to whatever you want.

Hi

Boris:
Since static_transform_publisher can be used only to set up transformation between fixed (w.r.t. each other) frames you have to use something like this in your node:

Well is use static_transform_publisher, because in my case odom -> map is static, only two 90° rotations.

I use this in my launch file:

<node pkg="tf" type="static_transform_publisher" name="odom_to_map"
    args="0.1 0  0  -1.570796327 0 -1.570796327  /map /odom 100" />

Search fpr static_transform_publisher in the ROS docu.

i installed 11.10 these days after i had ubuntu 12.04. since then my debugger information isn't as accurate as before, i don't know why. i use qtcreator. build type is debug.

well, the error message just tells me in which function i have the error, nothing more-

i have no idea why it isn't upper case here, but it is in qtcreator

so the data in /camera/depth/pointcloud with xyzrgb information is also not really synchron for positon and color ?

/camera/rgb/image_color and /camera/depth_registered/image at the same time synchron would be perfect

hi, i did that, but .. the two topics i subscribe publish in different frequencies. i need the rgb and d information as synchron as possible.

-------------------- SOLUTION ----------------- message_filters::TimeSynchronizer... has to be message_filters::Synchronizer... found that out after a day

How to mark as solved ?

Hi

I want to synchronize the image_color and the depth/image topic of a kinect.

both topics have different frequencies. i hoped before that they publish at the same time but sadly not.

is there a way to let these to topics been published simultaneously?

other posibility is to use a topic with rgb and d info, but i need an easy way to get the opencv rgb image.

the only topic i see, which has all info is "depth/points", but I need an opencv RGB image also, an iterative loop over the pointcloud to obtain the opencv image pixel by pixel seems to be studpi, because rgb image topic already exists.

why doesn'T that work:

private:
    message_filters::Subscriber<sensor_msgs::image> image_sub;
    message_filters::Subscriber<sensor_msgs::image> pointcloud_sub;

    message_filters::TimeSynchronizer<sensor_msgs::image, sensor_msgs::image=""> synchronizer;

public :

    // Constructor
    VisualAidingOpenni()
        :  nh("~")/*, it(nh)*/, synchronizer(1), ....

private:
void a function ()... {

        image_sub.subscribe(nh, topic_image, 1);
        pointcloud_sub.subscribe(nh, topic_pointcloud, 1);
        synchronizer.connectInput(image_sub, pointcloud_sub);
        synchronizer.registerCallback(boost::bind(&VisualAidingOpenni::subscriber_Callback,this , _1, _2));
}


void  subscriber_Callback(const sensor_msgs::ImageConstPtr& msg, const sensor_msgs::ImageConstPtr& pointcloud)
    {

...
}

forget the word pointcloud, tried the depth image at last, so everything in sensor_msgs::image

this doesn't even compile

private:
    message_filters::Subscriber<sensor_msgs::image> image_sub;
    message_filters::Subscriber<sensor_msgs::image> pointcloud_sub;
    typedef sync_policies::ApproximateTime<sensor_msgs::image, sensor_msgs::image=""> MySyncPolicy;
    message_filters::TimeSynchronizer<mysyncpolicy> synchronizer;

void function()
{
        image_sub = new message_filters::Subscriber(nh, topic_image, 1);
        pointcloud_sub = new message_filters::Subscriber(nh, topic_pointcloud, 1);
        synchronizer = new message_filters::TimeSynchronizer<mysyncpolicy>(MySyncPolicy(10), image_sub, pointcloud_sub);
 synchronizer.registerCallback(boost::bind(&VisualAidingOpenni::subscriber_Callback,this , _1, _2));

}

void  subscriber_Callback(const sensor_msgs::ImageConstPtr& msg, const sensor_msgs::ImageConstPtr& pointcloud)
    {

i get this error

/opt/hector/students/friedrich/hector_visual_aiding/hector_visaid_monocular/src/visual_aiding_openni.cpp:-1: In Elementfunktion »void VisualAidingOpenni::CamInfo_Callback(const CameraInfoConstPtr&)«:
:-1: Fehler:[CMakeFiles/visual_aiding_openni.dir/src/visual_aiding_openni.o] Fehler 1

THX

Doesn't work for me either. I also run Ubuntu 12.04 64bit fresh updated.