Dissertation
A weeding robot auto-leveling system for typical vegitable fields in pacific northwest region
Doctor of Philosophy (PhD), Washington State University
01/2019
Handle:
https://hdl.handle.net/2376/111368
Abstract
Weed control is a huge problem in organic vegetable farming. Even though there are machines available for inter-row weeding, manual weeding is the only choice for adequately controlling weeds in commercial production, particularly in the space between plants (also called intra-row weeding). Unfortunately, manual weeding is highly labor intensive and costly, and robotic weeding would be one of the potential solutions for solving this problem. However, there are some technical challenges to be overcome before robotic weeding could be practically implemented. The challenges mainly discussed in this research including three parts, 1) robot body posture change caused by uneven field surfaces in Pacific Northwest (PNW) region of USA; 2) weed detection in onion and carrot field in PNW region; 3) precise position control of the weeding end-effector for removing weeds located very close to the plant while the mobile robotic platform moving continuously on uneven field surfaces. To solve the challenges above, a self-leveling system was proposed and validated for ensuring end-effector positioning accuracy while operating on uneven terrain; a novel method of weed detection for robotic applications in onion and carrot fields typically found in Pacific Northwest (PNW) region of USA; and a dual-camera based end-effector position control system and assessing its performance in field operations. Validations tests in field conditions showed that when the robotic platform is travelling at a speed of 55 mm s-1 on a filed with a series of holes (20 cm deep and 30 cm apart), the self-leveling system could keep the angular error of leveling frame within 3ยบ; weed identification algorithm can achieve a true positive rate of over 99.7% and 99.2% respectively. False positive rate was less than 0.37% for onion and less than 0.83% for carrots. The processing time per image was less than 30 ms; this weeding robot could achieve an end-effector position control accuracy of less than 2 mm when the robot moving continuously on uneven field surface at a speed of 0.45 m s-1 which attained the design goal for this control system.
Metrics
16 File views/ downloads
54 Record Views
Details
- Title
- A weeding robot auto-leveling system for typical vegitable fields in pacific northwest region
- Creators
- Lin Chen
- Contributors
- Qin Zhang (Advisor)Manoj Karkee (Committee Member)Changki Mo (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Department of Biological Systems Engineering
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Number of pages
- 137
- Identifiers
- 99900581615401842
- Language
- English
- Resource Type
- Dissertation