PR2-RViz-Object-Recognition-Picking-Project icon indicating copy to clipboard operation
PR2-RViz-Object-Recognition-Picking-Project copied to clipboard

verified publisher icon WolfeTyler
→

Metadata

ROS project using object recognition using point-cloud data from an RGB-D camera to communicate pick & place instructions to a PR2 two-armed robot.

PR2 Robotic Object Recognition & Picking Project

Utilized Passthrough and RANSAC filtering on the point-cloud data alt text

Utilized Euclidean Clustering to distinguish identified objects for pick & place alt text

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 alt text

3 of 3 Objects Identified alt text

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

← Metadata

16
Stars
9
Forks
Watchers

Owner

verified publisher icon WolfeTyler

Metadata

ROS project using object recognition using point-cloud data from an RGB-D camera to communicate pick & place instructions to a PR2 two-armed robot.

Back