Medet's profile - activity

Hi, I am working on the same tutorial, and was trying to follow your steps, but have some problems:

1) where did you get your input_cloud? I am subscribing to kinect topic "/camera/depth/points", so my input cloud is of type sensor_msgs::PointCloud2ConstPtr&. How can I convert it to right type, as pcl::fromROSMsg(*input, input_cloud); giving me errors.

2) adding the bit you figured out giving other errors error: ‘sensor_msgs::PointCloud2_<std::allocator<void> >::Ptr’ has no member named ‘header’

I am quite new to ros & pcl, so any help will be much appreciated

Not really, I am running it on electric. My troubles were with recording process itself. Another issue is that specific object that I am trying to recognise, Nao robot, is bigger than any object used in training, that I have seen so far.

Any hints on packages/techniques to use for recognising such object?

Thanks for your response, but Roboearth was the first thing I started off, with no apparent success: couldn't manage recording process. Was looking for something less demanding to recording conditions

I am trying to create 3D models of objects for the 3D Object Recognition techniques presented in the following tutorial. I get the feeling that something similar to what I am looking for is done in the first steps of this tutorial but is glossed over.

I am wondering if there are any prebuilt packages either through PCL or ROS that cover this or if someone could point me in the best direction for getting started with doing this on my own.

My general thoughts were to use a rotating surface and to have it rotate the object while a kinect would capture information about the object and stitch together the various point clouds that would be generated.

Any input as always is more than welcome!