Release of PCL 0.8

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Release of PCL 0.8

Kent Williams
Hi all,

We are happy to announce the release of PCL 0.8

The 0.8 release brings:

  • flann

    • Updated to version 1.6.7
  • pcl

    • improved doxygen documentation overall
    • added support for indices for SampleConsensus models (previously broken)

    • fixed a grave bug in ProjectInliers, where row_step was not correctly set thus leading to (de)serialization issues in the new ros::Subscriber<PointCloud<T> > scheme

    • RangeImagePlanar can now also be created from a depth image.

    • Fixed a bug in RangeImagePlanar.

    • Fixed possible segmentation fault in RangeImageBorderExtractor.

    • RangeImage now has is_dense=false, since there typically are NANs in there.

    • Added Correspondence as a structure representing correspondences/matches (similar to OpenCV's DMatch) containing query and match indices as well as the distance between them respective points.
    • Added CorrespondenceEstimation for determining closest point correspondence, feature correspondences and reciprocal point correspondences.

    • Added CorrespondenceRejection and derivations for rejecting correspondences, e.g., based on distance, removing 1-to-n correspondences, RANSAC-based outlier removal (+transformation estimation).

    • Further splitted up registration.h and added transformation estimation classes, e.g., for estimating rigid transformation using SVD.
    • Added  sensor_msgs::Image image;  pcl::toROSMsg (cloud, image);, see tools/convert_pcd_image.cpp for a sample.

    • Added a new point type, PointWithScale, to store the output of SIFTKeypoint detection.

    • Fixed a minor bug in the error-checking in SIFTKeypoint::setScales(...)

    • Fixed small bug in MovingLeastSquares

    • Fixed small bug in GreedyProjectionTriangulation

    • Added unit tests for RSDEstimationMovingLeastSquares andGreedyProjectionTriangulation

    • Fixed and improved multiple point to line distance calculations, and added it to distance.h and unit testing
  • pcl_ros

    • Fixed a grave bug introduced in 0.7 where the NullFilter approach doesn't work

    • Implemented the exact time synchronizer path for PointCloudConcatenateDataSynchronizer

    • replaced subscribers/publishers in surface, segmentation, and feature nodelets fromsensor_msgs::PointCloud2 to pcl::PointCloud<T>, thus reducing complexity and getting rid of all the intermediate from/toROSMsg (de)serializations

    • moved away from storing state in the nodelets. The right way to do it is: callback -> PointCloud -> process -> publish, without storing the PointCloud as a member into the base class

    • introduced a mandatory emptyPublish which nodelets should use to send an empty result with the header frame_id and stamp equal with the input's.


      See the complete change-list at: www.ros.org/wiki/perception_pcl/ChangeList


       "We also released pcl_addons 0.3. See http://www.ros.org/wiki/perception_pcl_addons/ChangeList for more information"



      PS. CloudMan was proposed as the unofficial PCL mascot 

      (see http://pressthebuttons.typepad.com/photos/uncategorized/2008/09/26/cloudman_2.png for example)


--
Kent Williams

[hidden email]

_______________________________________________
[hidden email] / http://pointclouds.org
https://code.ros.org/mailman/listinfo/pcl-users
Reply | Threaded
Open this post in threaded view
|

Re: Release of PCL 0.8

Radu B. Rusu
Administrator
We forgot to mention: thanks to everyone that joined us on IRC and helped improve the doxygen documentation and
tutorials! Also big thanks to our Windows developers who spent many hours trying to rewrite parts of the code that were
Linux/MacOS specific.

The OpenCV team will start to work on PCL as early as next week too, and with their help, we'll be creating a PCL build
project that runs the unit tests on Windows, Linux, MacOS, for the PCL standalone version. This will help us catch and
fix errors before a release goes out. I believe we'll be using BuildBot for this (http://www.buildbot.net/).



For 0.9, we've started assembling a wishlist of things that should go into it. Please see
http://www.ros.org/wiki/pcl/Roadmap and feel free to add comments.

For 1.0, we'll start tackling the Registration/SLAM API, and form a task force that will work on that. There's at least
7-8 people that expressed interest in helping out on certain parts of the system, so that should give us critical mass
to come with a solid API. We aim to share the document that we've been prototyping the API on with all of you sometimes
next week.

Cheers,
Radu.
--
http://pointclouds.org

On 01/27/2011 11:54 PM, Kent Williams wrote:

> Hi all,
>
> We are happy to announce the release of PCL 0.8
>
> The 0.8 release brings:
>
>     *
>
>       flann <http://www.ros.org/wiki/flann>
>
>           o Updated to version 1.6.7
>     *
>
>       pcl <http://www.ros.org/wiki/pcl>
>
>           o improved doxygen documentation overall
>           o
>
>             added support for indices for SampleConsensus models (previously broken)
>
>           o
>
>             fixed a grave bug in ProjectInliers, where row_step was not correctly set thus leading to (de)serialization
>             issues in the new ros::Subscriber<PointCloud<T> > scheme
>
>           o
>
>             RangeImagePlanar can now also be created from a depth image.
>
>           o
>
>             Fixed a bug in RangeImagePlanar.
>
>           o
>
>             Fixed possible segmentation fault in RangeImageBorderExtractor.
>
>           o
>
>             RangeImage now has is_dense=false, since there typically are NANs in there.
>
>           o Added Correspondence as a structure representing correspondences/matches (similar to OpenCV's DMatch)
>             containing query and match indices as well as the distance between them respective points.
>           o
>
>             Added CorrespondenceEstimation for determining closest point correspondence, feature correspondences and
>             reciprocal point correspondences.
>
>           o
>
>             Added CorrespondenceRejection and derivations for rejecting correspondences, e.g., based on distance,
>             removing 1-to-n correspondences, RANSAC-based outlier removal (+transformation estimation).
>
>           o Further splitted up registration.h and added transformation estimation classes, e.g., for estimating rigid
>             transformation using SVD.
>           o
>
>             Added  sensor_msgs::Image image;  pcl::toROSMsg (cloud, image);, see tools/convert_pcd_image.cpp for a sample.
>
>           o
>
>             Added a new point type, PointWithScale, to store the output of SIFTKeypoint detection.
>
>           o
>
>             Fixed a minor bug in the error-checking in SIFTKeypoint::setScales(...)
>
>           o
>
>             Fixed small bug in MovingLeastSquares
>
>           o
>
>             Fixed small bug in GreedyProjectionTriangulation
>
>           o
>
>             Added unit tests for RSDEstimation, MovingLeastSquares andGreedyProjectionTriangulation
>
>           o Fixed and improved multiple point to line distance calculations, and added it to distance.h and unit testing
>     *
>
>       pcl_ros <http://www.ros.org/wiki/pcl_ros>
>
>           o
>
>             Fixed a grave bug introduced in 0.7 where the NullFilter approach doesn't work
>
>           o
>
>             Implemented the exact time synchronizer path for PointCloudConcatenateDataSynchronizer
>
>           o
>
>             replaced subscribers/publishers in surface, segmentation, and feature nodelets
>             fromsensor_msgs::PointCloud2 to pcl::PointCloud<T>, thus reducing complexity and getting rid of all the
>             intermediate from/toROSMsg (de)serializations
>
>           o
>
>             moved away from storing state in the nodelets. The right way to do it is: callback -> PointCloud -> process
>             -> publish, without storing the PointCloud as a member into the base class
>
>           o
>
>             introduced a mandatory emptyPublish which nodelets should use to send an empty result with the header
>             frame_id and stamp equal with the input's.
>
>
>             See the complete change-list at: www.ros.org/wiki/perception_pcl/ChangeList
>             <http://www.ros.org/wiki/perception_pcl/ChangeList>
>
>
>             "We also released pcl_addons 0.3. See http://www.ros.org/wiki/perception_pcl_addons/ChangeList for more
>             information"
>
>
>
>             PS. CloudMan was proposed as the unofficial PCL mascot
>
>             (see http://pressthebuttons.typepad.com/photos/uncategorized/2008/09/26/cloudman_2.png for example)
>
>
> --
> Kent Williams
>
> [hidden email] <mailto:[hidden email]>
>
>
>
> _______________________________________________
> [hidden email] / http://pointclouds.org
> https://code.ros.org/mailman/listinfo/pcl-users
_______________________________________________
[hidden email] / http://pointclouds.org
https://code.ros.org/mailman/listinfo/pcl-users