Date of Award

2-28-2024

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Mechanical, Automotive, and Materials Engineering

Keywords

3D Camera;Cucumber Characteristics;Image Detection;Robotic Arm;Shape and Colour Features;YOLO

Supervisor

Jill Urbanic

Abstract

Cucumbers constitute a significant portion of greenhouse vegetables cultivated in southern Ontario. The human labour shortage and potential injuries associated with manual labour in cucumber harvesting highlight the need to explore automation as a viable solution. The proposed automated system in this study comprises two key components: an image processing unit and a cutting robotic arm. The image processing phase involves the identification of cucumbers using six models using shape and colour features. Four models are successfully employed in YOLOv8, yielding results in the form of bounding boxes and keypoints, with no false positives. Near-real-time cucumber detection is also achieved using YOLO. One model has been successfully tested in RoboFlow. Finally, one new model is being initially investigated for the direct management of raw RGB-XYZ data in Python. About harvesting unit, a robotic arm equipped with a cutting tool is programmed to move to specific positions and perform cutting tasks. To enable this functionality, a 3D camera and cutter are integrated into the TCP of a UR3e robot. The 3D camera installed on TCP of the robot can move at angles and positions to capture cucumbers hidden from view. Additionally, analyses of cucumber geometry and force are carried out to improve the understanding of cucumber properties. Finally, initial cost and OEE analyses are conducted to assess the potential improvement resulting from transitioning from manual harvesting to automation.

Download

Share

COinS