Abstract
Accurate segmentation of pluckable zones and localization of plucking points are essential for intelligent tea harvesting. However, cultivar-specific traits, such as elongated internodes and low color contrast among buds, leaves, and stems, together with tea-shoot overlap and occlusion, hinder these tasks. To address these challenges, this study defines a pluckable zone for tea shoots and proposes a lightweight two-stage segmentation–localization pipeline. Stage I: an improved DeepLabV3+ model, termed DeepLabV3 + –MSEF, integrates a MobileNetV2 backbone, strip pooling–enhanced atrous spatial pyramid pooling, efficient channel attention, and focal loss for pluckable-zone segmentation. Stage II: two-dimensional (2D) plucking points are obtained from segmentation masks using a minimum bounding rectangle algorithm, and density-based spatial clustering of applications with noise (DBSCAN) is applied to the aligned three-dimensional (3D) point clouds to localize plucking points amid shoot overlap and occlusion. On an eight-cultivar test set, DeepLabV3 + -MSEF features 6.04 M parameters, runs at 60.14 frames per second (FPS), and achieves a mean intersection over union (mIoU) of 81.77% and an F1-score of 89.12%. Compared with the original DeepLabV3+ with an Xception backbone, it improves mIoU by 10.78 percentage points, reduces the parameter count from 54.71 M to 6.04 M, and nearly doubles inference speed from 30.36 FPS to 60.14 FPS, while achieving a favorable accuracy–efficiency tradeoff among benchmark models. Cultivar-wise testing demonstrates stable segmentation performance, with the mIoU ranging from 75.36% to 85.26% and F1-scores from 84.71% to 91.62%; residual errors mainly stem from variable illumination, occlusion by mature or fish leaves, and low color contrast among buds, leaves, and stems. The 2D localization stage achieves a cultivar-weighted success rate of 85.17% on 400 field images, while DBSCAN-based 3D point-cloud processing yields 92.00% plucking-point accuracy in occluded and overlapping shoot scenarios. These results demonstrate the potential of the proposed method for real-time, cross-cultivar intelligent tea harvesting.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by-nc-nd. 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 12
Claim · model-written summary
The model used 6.04 M parameters.
Verbatim
e0“DeepLabV3 + -MSEF features 6.04 M parameters”
t0“tea shoots”
Claim · model-written summary
The model ran at 60.14 frames per second.
Verbatim
e0“runs at 60.14 frames per second (FPS)”
t0“tea shoots”
Claim · model-written summary
Mean intersection over union was 81.77%.
Verbatim
e0“mean intersection over union (mIoU) of 81.77%”
t0“tea shoots”
Claim · model-written summary
F1 score was 89.12%.
Verbatim
e0“F1-score of 89.12%”
t0“tea shoots”
Claim · model-written summary
Against original DeepLabV3+, mIoU improved by 10.78 percentage points.
Verbatim
e0“Compared with the original DeepLabV3+ with an Xception backbone, it improves mIoU by 10.78 percentage points”
t0“tea shoots”
Claim · model-written summary
Parameter count fell from 54.71 M to 6.04 M.
Verbatim
e0“reduces the parameter count from 54.71 M to 6.04 M”
t0“tea shoots”
Claim · model-written summary
Inference speed rose from 30.36 FPS to 60.14 FPS.
Verbatim
e0“nearly doubles inference speed from 30.36 FPS to 60.14 FPS”
t0“tea shoots”
Claim · model-written summary
Cultivar mIoU ranged from 75.36% to 85.26%.
Verbatim
e0“mIoU ranging from 75.36% to 85.26%”
t0“tea shoots”
Claim · model-written summary
Cultivar F1 scores ranged from 84.71% to 91.62%.
Verbatim
e0“F1-scores from 84.71% to 91.62%”
t0“tea shoots”
Claim · model-written summary
Residual errors mainly came from illumination, occlusion and color contrast.
- Taxon
- tea shootst0
- Outcome
- residual errorse0
- Qualifier
- variable illumination, occlusion by mature or fish leaves, and low color contrast among buds, leaves, and stemse0
Not reportedComparator
Verbatim
e0“residual errors mainly stem from variable illumination, occlusion by mature or fish leaves, and low color contrast among buds, leaves, and stems”
t0“tea shoots”
Claim · model-written summary
Two dimensional localization had an 85.17% cultivar weighted success rate on 400 field images.
Verbatim
e0“2D localization stage achieves a cultivar-weighted success rate of 85.17% on 400 field images”
t0“tea shoots”
Claim · model-written summary
DBSCAN based three dimensional processing achieved 92.00% plucking point accuracy in occluded and overlapping shoots.
Verbatim
e0“DBSCAN-based 3D point-cloud processing yields 92.00% plucking-point accuracy in occluded and overlapping shoot scenarios”
t0“tea shoots”