ColorizeRandomForest

What Is This

../../_images/colorize_random_forest.png

Extract FPFH (Fast Point Feature Histograms) from input point cloud, and classify each cluster into 2 categories using Random Forest classifier.

This node internally applies x, y and z filter and euclidean clustering as preprocessing.

Subscribing Topic

  • ~input (sensor_msgs/PointCloud2)

    Input point cloud from which specified part will be extracted.

Publishing Topic

  • ~output/zero (sensor_msgs/PointCloud2)

    Output point cloud whose classification result == 0

  • ~output/nonzero (sensor_msgs/PointCloud2)

    Output point cloud whose classification result != 0

Parameter

  • ~rs (Float, default: 0.03)

    Radius in meters used for searching FPFH.

  • ~po (Float, default: 0.03)

    Offset in meters used for pass through filter.

  • ~po2 (Float, default: 0.06)

    Another offset in meters used for pass through filter.

  • ~sum_num (Int, default: 100)

    Number of points used for averaging FPFH histogram.

Calling Service

  • ~classify_server (ml_classifiers/ClassifyData)

    Random Forest server for classification.

Sample

roslaunch jsk_pcl_ros sample_colorize_random_forest.launch