Abstract
Aimed to improve the quality of picked tea leaves and the efficiency of tea harvesting, an adaptive oolong tea harvesting robot with an adjustment module of a cutting tool and a harvesting line localization algorithm is proposed. The robot includes a vision measurement module and an adjustment mechanism of a cutting tool, enabling it to assess the shape of tea bushes and adaptively adjust the cutter configuration. To address the challenges of complex tea bush structures and environmental noise, a Prior–Tukey RANSAC algorithm was proposed for accurate harvesting model fitting. Our algorithm leverages prior knowledge about tea bush stem characteristics, uses the Tukey loss function to enhance robustness to outliers, and incorporates workspace constraints to ensure that the cutting tool remains within feasible operational limits. To evaluate the performance of the robot, experiments were conducted in a tea garden in Wuyi Mountain, China. Under ideal conditions, our algorithm achieved an inlier ratio of 43.10% and an R2 value of 0.9787, significantly outperforming traditional RANSAC and other variants. Under challenging field conditions, the proposed algorithm demonstrated robustness, maintaining an inlier ratio of 47.50% and an R2 value of 0.9598. And the processing time of the algorithm met the real-time requirements for effective tea-picking operations. The field experiments also showed an improvement in intact tea rates, from 79.34% in the first harvest to 81.57% in the second harvest, with a consistent usable tea rate of around 85%. Additionally, the robot had a harvesting efficiency of 260.14 kg/h, which was superior to existing handheld and riding-type tea pickers. These results indicate that the robot effectively balances efficiency, accuracy, and robustness, providing a promising solution for high-quality tea harvesting in complex environments.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 8
Claim · model-written summary
Under ideal conditions the algorithm reached 43.10% inliers.
Verbatim
e0“Under ideal conditions, our algorithm achieved an inlier ratio of 43.10% and an R2 value of 0.9787”
t0“Tea”
Claim · model-written summary
Under ideal conditions model fit reached R2 0.9787 and outperformed RANSAC variants.
Verbatim
e0“an R2 value of 0.9787, significantly outperforming traditional RANSAC and other variants”
t0“Tea”
Claim · model-written summary
Under challenging field conditions the algorithm maintained 47.50% inliers.
Verbatim
e0“Under challenging field conditions, the proposed algorithm demonstrated robustness, maintaining an inlier ratio of 47.50% and an R2 value of 0.9598”
t0“Tea”
Claim · model-written summary
Under challenging field conditions model fit was R2 0.9598.
Verbatim
e0“an R2 value of 0.9598”
t0“Tea”
Claim · model-written summary
Algorithm processing time met real-time tea picking requirements.
Verbatim
e0“the processing time of the algorithm met the real-time requirements for effective tea-picking operations”
t0“Tea”
Claim · model-written summary
Intact tea rate improved from 79.34% in the first harvest to 81.57% in the second.
Verbatim
e0“an improvement in intact tea rates, from 79.34% in the first harvest to 81.57% in the second harvest”
t0“Tea”
Claim · model-written summary
Usable tea rate stayed around 85%.
Verbatim
e0“a consistent usable tea rate of around 85%”
t0“Tea”
Claim · model-written summary
Robot harvesting efficiency was 260.14 kg/h, above handheld and riding-type pickers.
Verbatim
e0“the robot had a harvesting efficiency of 260.14 kg/h, which was superior to existing handheld and riding-type tea pickers”
t0“Tea”