---

PR2 Robotic Object Recognition & Picking Project

Utilized Passthrough and RANSAC filtering on the point-cloud data

Utilized Euclidean Clustering to distinguish identified objects for pick & place

roslaunch sensor_stick training.launch
rosrun sensor_stick capture_features.py

Output is training_set.sav file

rosrun sensor_stick train_svm.py

Output is Confusion Matrices and model.sav file

Test World 1

Confusion Matrix - Not Confused

3 of 3 Objects Identified

Output_1 YAML file included

Links

• PCL documentation : http://strawlab.github.io/python-pcl/
• RANSAC algorithm : http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/FISHER/RANSAC/
• Outlier Removal (paper) : http://people.csail.mit.edu/changil/assets/point-cloud-noise-removal-3dv-2016-wolff-et-al.pdf
• Clustering Algorithm : http://bit.ly/clustering-tutorial
• Segmentation with NN (intro) : http://bit.ly/segmentation-intro-nn