Identification of Pruning Branches for for Automated Dormant - PowerPoint PPT Presentation

Identification of Pruning Branches for for Automated Dormant Pruning M Manoj Karkee j K k Assistant Professor Biological Systems Engineering Department Biological Systems Engineering Department Center for Precision and Automated Agricultural Systems Washington State University

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 PRUNING • Removal of unwanted branches from Removal of unwanted branches from trees • Improves fruit quality and tree health Improves fruit quality and tree health • Maintains canopy shape and size • Labor intensive and laborious • ~30% of pre-harvest cost in apples 30% o p e a est cost app es

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012 Simpler tree architectures open up Simpler tree architectures open up opportunity for automated pruning

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 LONG- TERM GOAL Automated Selective Pruning g SPECIFIC OBJECTIVE SPECIFIC OBJECTIVE Identify Branches for Automated Pruning

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 STEPS • Study Manual Pruning Study Manual Pruning • Reconstruct 3D Skeleton of Trees • Identify Pruning Branches Identify Pruning Branches • Compare with Manual Pruning

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 STUDY SITE • Alan Bros. Inc, Prosser, WA • Tall Spindle Architecture • Inter-row – 2.44m, inter-plant - 0.45m Inter row 2.44m, inter plant 0.45m • Fourth Year

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012 HUMAN PRUNING EVALUATION

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012 HUMAN PRUNING EVALUATION

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 WOOD REMOVAL BY WORKERS WOOD REMOVAL BY WORKERS

Interaction with Horticulturists and In Addition Growers

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 OUTCOME Pruning is quite subjective; but some Pruning is quite subjective; but some common guidelines can be formulated!! Two Guidelines • Remove long branches Remove long branches • Remove old/large branches • Remove closely spaced branches • Remove overlapping branches

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 Pruning Branch ID: Functional Block Diagram I Image Image I Skeleton Sk l Skeletonization Acquisition Preprocessing Analysis

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 IMAGE ACQUISITION 3D Camera Pan-Tilt System

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 PREPROCESSING Noise Filtering Stitching

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 SKELETONIZATION SKELETONIZATION

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 PRUNING BRANCH IDENTIFICATION Remove Remove • Long branches • Closely spaced Closely spaced branches

IEEE RAS Ag Robotics and Automation Webinar May 6, 2013 SUMMARY Based on branch length and spacing criteria • Achieved 86% accuracy in identifying pruning branches • Percentage wood removal was within 2% of most experience worker Video http://www.youtube.com/watch?v=s12YytfKDU4

Specialty Crop Engineering Solutions Workshop Pittsburgh, PA; Nov 28, 2012 ONGOING AND FUTURE EFFORTS • Pruning Point Localization • Computational Optimization Computational Optimization • End Effector Evaluation • Robotic System Development • Robotic System Development

Thank You Manoj Karkee manoj.karkee@wsu.edu