segmenters_lib

　　The LiDAR segmenters library, for segmentation-based detection.

How to use

1. We name your ros workspace as CATKIN_WS and git clone as a ROS package, with common_lib and object_builders_lib as dependencies.

   $ cd $(CATKIN_WS)
   # we recommand you to organize your workspace as following
   $ mkdir -p src/common
   $ mkdir -p src/perception/libs
   
   # git clone basic libraries, like common_lib
   $ cd $(CATKIN_WS)/src/common
   $ git clone https://github.com/LidarPerception/common_lib.git libs
   
   # git clone perception libraries, segmenters_lib and its dependencies
   $ cd $(CATKIN_WS)/src/perception/libs
   $ git clone https://github.com/LidarPerception/roi_filters_lib.git roi_filters
   $ git clone https://github.com/LidarPerception/object_builders_lib.git object_builders
   $ git clone https://github.com/LidarPerception/segmenters_lib.git segmenters
   
   # build your ros workspace for our segmentation-based detection demo
   # uncomment add_subdirectory() in src/perception/libs/segmenters/CMakeLists.txt, git diff as following:
   -#add_subdirectory(example)
   +add_subdirectory(example)
   $ cd $(CATKIN_WS)
   $ catkin build -DCMAKE_BUILD_TYPE=Release
2. Run demo under KiTTI raw dataset using kitti_ros's replayer.

   $ cd $(CATKIN_WS)/src
   $ git clone https://github.com/LidarPerception/kitti_ros.git
   # build your ros workspace for our segmentation-based detection demo
   $ cd ..
   $ catkin build -DCMAKE_BUILD_TYPE=Release
   • Terminal 1: KiTTI raw dataset replay, more tutorials.

     $ cd $(CATKIN_WS)
     $ source devel/setup.bash
     # change Mode for Keyboard Listening Device
     $ sudo chmod 777 /dev/input/event3
     # launch kitti_ros's kitti_player for frame-by-frame algorithm testing
     $ roslaunch kitti_ros kitti_player.launch
   • Terminal 2: launch Seg-based Detector demo.

     $ cd $(CATKIN_WS)
     $ source devel/setup.bash
     $ roslaunch segmenters_lib demo.launch
3. Follow the demo example to use our LiDAR segmenters library.
   • Cascadingly use roi_filter, ground_remover and non_ground_segmenter for Point Cloud perception, like our Seg-based Detector: detection_node.
   • Refer to our CMakeLists.txt for building your own ros package using this library.

   

Parameters

• Demo parameters, defined in detection.yaml
  • Subscribe Point Cloud in topic sub_pc_topic, default is /kitti/points_raw (sensor_msgs/PointCloud2).
  • Publish Ground Point Cloud in topic pub_pc_ground_topic, default is /segmenter/points_ground (sensor_msgs/PointCloud2).
  • Publish Non-Ground Point Cloud in topic pub_pc_nonground_topic, default is /segmenter/points_nonground (sensor_msgs/PointCloud2).
  • Publish Candidate Objects Cloud, different objects with different intensity, in topic pub_pc_clusters_topic, default is /segmenter/points_clustered (sensor_msgs/PointCloud2).
• Segmenters algorithm parameters, defined in segmenter.yaml.
  • Get these paramters using common::getSegmenterParams(const ros::NodeHandle& nh, const std::string& ns_prefix) in common_lib.
  • Adjust some paramters to your hardware installation.

TODO lists

Ground Segmenters

• PCL RANSAC, ICCV2011 PCL-Segmentation

  Refer: PCL: Plane model segmentation

• GPF (Ground Plane Fitting), ICRA 2017

  @inproceedings{zermas2017fast,
    title={Fast segmentation of 3d point clouds: A paradigm on lidar data for autonomous vehicle applications},
    author={Zermas, Dimitris and Izzat, Izzat and Papanikolopoulos, Nikolaos},
    booktitle={Robotics and Automation (ICRA), 2017 IEEE International Conference on},
    pages={5067--5073},
    year={2017},
    organization={IEEE}
  }
• linefit_ground_segmentation, IV 2010

  @inproceedings{himmelsbach2010fast,
    title={Fast segmentation of 3d point clouds for ground vehicles},
    author={Himmelsbach, Michael and Hundelshausen, Felix V and Wuensche, H-J},
    booktitle={Intelligent Vehicles Symposium (IV), 2010 IEEE},
    pages={560--565},
    year={2010},
    organization={IEEE}
  }
• depth_clustering, IROS 2016

  @inproceedings{bogoslavskyi2016fast,
    title={Fast range image-based segmentation of sparse 3D laser scans for online operation},
    author={Bogoslavskyi, Igor and Stachniss, Cyrill},
    booktitle={Intelligent Robots and Systems (IROS), 2016 IEEE/RSJ International Conference on},
    pages={163--169},
    year={2016},
    organization={IEEE}
  }
  
• Scan Line Run, ICRA 2017, like Standford's Junior self-driving car.

  @inproceedings{zermas2017fast,
    title={Fast segmentation of 3d point clouds: A paradigm on lidar data for autonomous vehicle applications},
    author={Zermas, Dimitris and Izzat, Izzat and Papanikolopoulos, Nikolaos},
    booktitle={Robotics and Automation (ICRA), 2017 IEEE International Conference on},
    pages={5067--5073},
    year={2017},
    organization={IEEE}
  }
  @incollection{montemerlo2009junior,
    title={Junior: The stanford entry in the urban challenge},
    author={Montemerlo, Michael and Becker, Jan and Bhat, Suhrid and Dahlkamp, Hendrik and Dolgov, Dmitri and Ettinger, Scott and Haehnel, Dirk and Hilden, Tim and Hoffmann, Gabe and Huhnke, Burkhard and others},
    booktitle={The DARPA Urban Challenge},
    pages={91--123},
    year={2009},
    publisher={Springer}
  }
• Deep-learning: FCN, IV 2017

  @inproceedings{caltagirone2017fast,
    title={Fast LIDAR-based road detection using fully convolutional neural networks},
    author={Caltagirone, Luca and Scheidegger, Samuel and Svensson, Lennart and Wahde, Mattias},
    booktitle={Intelligent Vehicles Symposium (IV), 2017 IEEE},
    pages={1019--1024},
    year={2017},
    organization={IEEE}
  }

Non-ground Segmenters

• PCL Euclidean Cluster Extraction, ICCV2011 PCL-Segmentation

  Refer: PCL: Euclidean Cluster Extraction

• Region-based Euclidean Cluster Extraction

  @inproceedings{yan2017online,
    title={Online learning for human classification in 3d lidar-based tracking},
    author={Yan, Zhi and Duckett, Tom and Bellotto, Nicola},
    booktitle={Intelligent Robots and Systems (IROS), 2017 IEEE/RSJ International Conference on},
    pages={864--871},
    year={2017},
    organization={IEEE}
  }
• Model-based Segmentation, ITSC 2017

  @article{shin2017real,
    title={Real-time and accurate segmentation of 3-D point clouds based on Gaussian process regression},
    author={Shin, Myung-Ok and Oh, Gyu-Min and Kim, Seong-Woo and Seo, Seung-Woo},
    journal={IEEE Transactions on Intelligent Transportation Systems},
    volume={18},
    number={12},
    pages={3363--3377},
    year={2017},
    publisher={IEEE}
  }
• Probabilistic Framework, RSS 2016

  @inproceedings{held2016probabilistic,
    title={A Probabilistic Framework for Real-time 3D Segmentation using Spatial, Temporal, and Semantic Cues.},
    author={Held, David and Guillory, Devin and Rebsamen, Brice and Thrun, Sebastian and Savarese, Silvio},
    booktitle={Robotics: Science and Systems},
    year={2016}
  }

  

• Tracking-help Segmentation. IV, 2012. Implemented in our tracking_lib.

  @inproceedings{himmelsbach2012tracking,
    title={Tracking and classification of arbitrary objects with bottom-up/top-down detection},
    author={Himmelsbach, Michael and Wuensche, H-J},
    booktitle={Intelligent Vehicles Symposium (IV), 2012 IEEE},
    pages={577--582},
    year={2012},
    organization={IEEE}
  }