This repository includes classical segmentation methods for Point-Clouds (PCs) implemented as a Robotic-Operating-System (ROS) node. The implemented segmentation methods are:
- Euclidean Cluster Extraction,
- Region Growing Segmentation,
- Plane model Segmentation,
- Supervoxel Segmentation.
Authors: Jens Lundell and Francesco Verdoja
Maintainer: Jens Lundell, jens.lundell@aalto.fi
Affiliation: Intelligent Robotics Lab, Aalto University
Fast 3D Pointcloud Segmentation
Clone or download the project from Github:
cd <PATH_TO_YOUR_CATKIN_WORKSPACE>/src
git clone git@github.com:aalto-intelligent-robotics/point_cloud_segmentation.git
Compile the ROS workspace
cd <PATH_TO_YOUR_CATKIN_WORKSPACE>
catkin_make
The simplest option to run the segmentation node is to do the following
roslaunch point_cloud_segmentation segmentation.launch
First start the Kinect camera
roslaunch openni_launch openni.launch
Then launch the segmentation node
roslaunch point_cloud_segmentation segmentation.launch segmentation_algorithm="region"
Finally run rviz
roscd point_cloud_segmentation
rosrun rviz rviz -d rviz/point_cloud_segmentation_config.rviz
point_cloud_segmentation takes a point-cloud as input and generates a segmented point-cloud. The node uses lazy publishing and will not publish the segmented point-cloud until there is a subscriber for point_cloud_segmentation/segmented_pc.
~pc_topic (sensor_msgs/PointCloud, default: /camera/depth_registered/points) - The input point-cloud topic.
~camera_frame (String, default camera_frame) - Name of the camera frame used.
~base_frame (String. default base_link) - Name of the base frame that the point-cloud is transformed into for doing background subtraction. One optioni for setting the base_frame is to estimate the plane where the objects are lying with an Aruco marker.
~x_cutoff ([float, float], default [-0.3, 0.3]) - Points outside this region in x-direction are removed.
~y_cutoff ([float, float], default [-0.3, 0.3]) - Points outside this region in y-direction are removed.
~z_cutoff ([float, float], default [0.001, 0.3]) - Points outside this region in z-direction are removed.
~segmentation_algorithm (String, default base) - The segmentation method to use. Base method only does background subtraction. The other alternatives are: plane, eucledian, region, superclustering.
~inlier_threshold (float, default 0.01) - Determines how close a point must be to the model in order to be considered an inlier.
~leaf_size (float, default 0.005) - Specifies the down-sampling rate of the point-cloud when using the Eucledian Cluster Segmentation method.
~min_cluster_size (int, default 300) - The minimum number of points a cluster need to have to be considered a segment. Only used for Eucledian Cluster Segmentation and Region Growing Segmentation.
~max_cluster_size (int, default 100000) - The maximum number of points a cluster can have to be considered a segment.
~cluster_tolerance (float, default 0.01) - the maximum distance a point may have to a cluster member to also be counted as member of that cluster. Only used for Eucledian cluster segmentation.
~curvature_threshold (float, default 1) - The maximum deviation between the normals of two points to be in the same cluster.
~smoothness_threshold (float, default 0.4) - The maximum deviation between point normals to be in the same cluster. Only used for Region Growing Segmentation.
~number_of_neighbours (int, default 80) - number of neighbors to use for the K nearest search. Only used for Region Growing Segmentation.
~voxel_resolution (float, default 0.005) - Sets the voxel size, which determines the leaf size of the underlying octree structure (in meters). Only used for Supervoxel Segmentation.
~seed_resolution (float, default 0.1) - Sets the seeding size, which determines how big the supervoxels will be (in meters). Only used for Supervoxel Segmentation.
~color_importance (float, default 0.2) - Sets the weight for color - how much color will influence the shape of the supervoxels. Only used for Supervoxel Segmentation.
~spatial_importance (float, default 0.4) - Sets the weight for color - how much color will influence the shape of the supervoxels. Only used for Supervoxel Segmentation.
~normal_importance (float, default 1.0) - Sets the weight for normal - how much surface normals will influence the shape of the supervoxels. Only used for Supervoxel Segmentation.
~threshold. (float, 0.5) - Determines which level of the dendogram represents the best segmentation. Only used for Supervoxel Segmentation.
~rgb_color_space (bool, default false) - Uses the RGB color space for measuring the color distance; if false, LAB* color space is used. Only used for Supervoxel Segmentation.
~convexity_criterion (bool, default true) - Uses the convexity criterion to weigh the geometric distance; if false, convexity is not considered. Only used for Supervoxel Segmentation.
~adapt_lambda (bool, default false) - Uses Adaptive lambda as merging criterion. Only used for Supervoxel Segmentation.
~equalization (bool, default false) - Uses Equalization as merging criterion; if false, 200 bins are used. Only used for Supervoxel Segmentation.
pc_topic (sensor_msgs/PointCloud) - The topic for the input point-cloud. This is typically set to /camera/depth_registered/points if you use a Kinect camera.
point_cloud_segmentation/segmented_pc (sensor_msgs/PointCloud) - The output segmented point-cloud color coded according to the segments.
point_cloud_segmentation/scene_segmentation (input: sensor_msgs/PointCloud, output: point_cloud_segmentation/SegmentedPointCloud) - This sevice takes a point-cloud as input and returns a segmented point-cloud. The custom return message contains the original point-cloud color coded according to the segments, a list of the segmented point-clouds, and a list of indices representing which points belong to which segment.
point_cloud_segmentation/change_segmentation_method (input: String, output: bool) - This service lets the user change the segmentation method at run-time. It takes as input a string that should be one of the following plane, eucledian, region, or superclustering and returns a boolean indicating if the change was successful or not.
point_cloud_segmentation/update_base_params - This service lets the user update the x_cutoff, y_cutoff, and z_cutoff parameters.
point_cloud_segmentation/update_base_params - This service lets the user update the x_cutoff, y_cutoff, and z_cutoff parameters.
point_cloud_segmentation/update_params - This service lets the user update the parameters of the currently used segmentation algorithm.
This project is licensed under the MIT License - see the LICENSE file for details