News Articles

PCL meetups

The recent ICRA (International Conference on Robotics and Automation) research conference hosted another successful PCL tutorial on Friday, May 10. Given the fact that some of the PCL luminaries were present at the event,  this was the perfect opportunity to use the day prior to the event and hold a meeting for a generic "PCL roadmap" discussion.

 

Taken almost as a scene from a Cosa Nostra movie, the group talked about PCL 1.7, interoperability issues regarding the PointCloud2 format, but also the way to push forward agendas using pcl-developers@ as a forum for discussions. Decisions such as "to break" or not a part of the API have always been open for vote in PCL, and as the code base and the community continue to grow, these decisions and discussions around them are going to be increasingly more important.

 

Such meetings have been proven to be very beneficial in the past as well for PCL, with all sorts of different collaborations being spawn between participants, some leading to joint research publications, while others resulting in better code contributions. Given that there's a plethora of scientific events being organized each year, we thought about encouraging any developers to propose such single day PCL tutorial sessions, in order to get a chance to meet your fellow colleague developers, but more importantly your users.…

Kinect for Windows with PCL

PCL's actual grabber interface provides a smooth and convenient access to different devices and their drivers, file formats and other sources of data. The first driver that was incorporated is the OpenNI Grabber, which makes it a breeze to request data streams from OpenNI compatible cameras. The cameras that have been tested so far are the Primesense Reference Design, Microsoft Kinect for XBox and Asus Xtion Pro.

Microsofts Kinect for Windows was not compatible with PCL yet, because it is driven by the Kinect SDK which is not compatible with OpenNI.

problem.jpg

Now there is an Open Source package available containing a dll module which functions as bridge between the Kinect SDK, OpenNI and PCL.

bridge_solution.jpg oni_recording.jpg oni_playback.jpg

Project URL: https://github.com/mdkus/kinect-mssdk-openni-pcl-bridge/

This bridge module was developped by Tomoto S. Washio and made compatible to PCL by Michael Dingerkus. For compatibility there is a patch included to install extensions for PCL in terms of handling…

PCL-ORCS - code sprint success!

Ocular Robotics        Open Perception

Pat Marion, an R&D Engineer at Kitware, has concluded his work on the joint Open Perception - Ocular Robotics code sprint.  The goal of the code sprint was to develop a pcl::Grabber driver interface for the  RE0x laser sensors and to develop visualization code capable of real-time display of point cloud data streams acquired by the grabber interface.

The pcl::RobotEyeGrabber is a new C++ class that was added to the pcl io module.  It implements the pcl::Grabber interface and provides access to point cloud data streams sent over the network by RE0x laser sensors.  A point cloud visualization application called RobotEye Viewer was developed using the RobotEyeGrabber and PCLVisualizer.  The application uses the RobotEye C++ API to send control messages to the RobotEye laser sensor and displays real-time point cloud streams acquired using the grabber interface.

The following video demonstrates the RobotEye Viewer application in action using a RE05 laser sensor:

 

Additional…


SwRI and NIST sponsor a new PCL code sprint

The Southwest Research Institute (SwRI) and National Institute of Standards and Technology (NIST) are sponsoring a new PCL code sprint! The efforts will be focused on developing algorithms for human detection and tracking, out of 2D camera imagery fused with 3D point cloud data (coming from ASUS XtionPRO cameras).

Interested candidates should submit the following information to jobs@pointclouds.org:

  • a brief resume
  • a list of existing PCL contributions (if any)
  • a list of projects (emphasis on open source projects please) that they contributed to in the past

This project requires good C++ programming skills, and knowledge of PCL internals.

3 new code sprints from Toyota and Open Perception

Toyota has been a long term supporter of PCL, and pretty much created the concept of code sprints for PCL, with the first PCL code sprint ever: TOCS! This year, we have partnered with our colleagues from Toyota again for a series of 3 new exciting code sprint projects:

  • Primitive shape (cylinders, spheres, cones, etc.) recognition in point cloud data
  • Segmentation/Clustering of objects in cluttered environments
  • 3D feature development and benchmarking

PCL-TOCS(#2) will run for 3 months during the spring of 2013. As always, interested candidates should submit the following information to jobs@pointclouds.org:

  • a brief resume
  • a list of existing PCL contributions (if any)
  • a list of projects (emphasis on open source projects please) that they contributed to in the past

This project requires good C++ programming skills, and knowledge of PCL internals.

Spectrolab PCL code sprint

It is our pleasure to announce a new code sprint from our host organization, Open Perception, and Spectrolab, a Boeing company. Spectrolab has expressed interest in connection with the potential possibilities that PCL offers, and we will be searching for outstanding candidates to participate in a new code sprint that involves the development of 3D Viewer Software in PCL, as well as an advanced sensor grabber for Spectrolab's SpectroScan3D LIDAR imager.

The sprint will run for 3 months in the spring of 2013. Potential candidates should submit the following information to jobs@pointclouds.org:

  • a brief resume
  • a list of existing PCL contributions (if any)
  • a list of projects (emphasis on open source projects please) that they contributed to in the past

This project requires good C++ programming skills, knowledge of PCL internals and a basic understanding of laser sensors and 3D visualization.

Basic PCD Matlab interface available

A simple interface to MATLAB is available at 
 
 
It includes pure MATLAB code to read and write unorganized point clouds as PCD files and a wrapper function for point cloud visualization that writes the MATLAB data to a temporary file and sets pcl_viewer loose on it.  Using files is inelegant and inefficient, but we sidestep the whole problem of trying to create MEX files linked to PCL.
 
This is a work in progress and things to add are writing binary PCD; reading/writing binary_compressed formatted files; handling organized data.
 
 

New code sprints from Honda Research Institute

After a successful first code sprint with HRI, it is our pleasure to announce the beginning of 4 new projects:

  1. labeling outdoor pedestrian and car data as ground truth (2 months)
  2. fast 3D cluster recognition of pedestrians and cars in uncluttered scenes (3 months)
  3. part-based 3D recognition of pedestrians and cars in cluttered scenes (6 months)
  4. stereo-based road area detection (2 months)

Please see the previous HRCS sprint announcement for more information. PCL-HRCS will run for 3-6 months during Q1 2013. Interested candidates should submit the following information to jobs@pointclouds.org:

  • a brief resume
  • a list of existing PCL contributions (if any)
  • a list of projects (emphasis on open source projects please) that they contributed to in the past

This project requires good C++ programming skills, and knowledge of PCL internals.

Python bindings for the Point Cloud Library

We are proud to to announce the release of python-pcl Python bindings for PCL.

Now you can use the power and performance of PCL from the comfort of Python. Currently the following features of PCL, using PointXYZ point clouds, are available;

  • I/O and integration; saving and loading PCD files
  • segmentation
  • sample consensus model fittting (RANSAC + others, cylinders, planes, common geometry)
  • smoothing (median least squares)
  • filtering (voxel grid downsampling, passthrough, statistical outlier removal)
  • exporting, importing and analysing pointclouds with numpy

An simple demonstration showing the statistical outlier filter:

import pcl
p = pcl.PointCloud()
p.from_file("table_scene_lms400.pcd")
fil = p.make_statistical_outlier_filter()
fil.set_mean_k (50)
fil.set_std_dev_mul_thresh (1.0)
fil.filter().to_file("inliers.pcd")

For a more complete example showing how to combine filtering, plane and cylinder segmentation (the code used to generate the logo above), see this example.

For more information please see the examples, tests,…

Leica Geosystems partnership

 

It is our immense pleasure to announce the beginning of a new partnership between Open Perception and Leica Geosystems, and a series of new code sprints: PCL-LGSCS!

We are looking for talented contributors that are willing to develop open source efficient compression mechanisms for organized and unorganized 3D point cloud data in two separate code sprint projects. In both cases, the ultimate goal is to achieve as much compression as possible, but we will require an analysis with respect to the de/compression speed that can be obtained. A requirement is to be able to process a few million points per second on a standard laptop. In addition we will analyze and compare different lossy vs lossless compression techniques.

The data sources are variate, from terrestrial to mobile and aerial point clouds. Besides XYZ coordinates we expect the datasets to contain an additional intensity and/or color per point. Each dataset will contain standard meta information, and in the case of lossy compression, we will need to specify certain error limits to be satisfied.

The organized data format consists of a series…

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • >>
  • Last
  • Latest Posts

    Latest Comments