Fresh Tea Leaf-Grading Detection: An Improved YOLOv8 Neural Network Model Utilizing Deep Learning DOI Creative Commons
Zejun Wang, Yuxin Xia,

Houqiao Wang

et al.

Horticulturae, Journal Year: 2024, Volume and Issue: 10(12), P. 1347 - 1347

Published: Dec. 15, 2024

To facilitate the realization of automated tea picking and enhance speed accuracy leaf grading detection, this study proposes an improved YOLOv8 network for fresh recognition. This approach integrates a Hierarchical Vision Transformer using Shifted Windows to replace segments original YOLOv8’s architecture, thereby alleviating computational load dense image processing tasks reducing expenses. The incorporation Efficient Multi-Scale Attention Module with Cross-Spatial Learning serves attenuate influence irrelevant features in complex backgrounds, which turn, elevates model’s detection Precision. Additionally, substitution loss function SIoU facilitates more rapid model convergence precise pinpointing defect locations. empirical findings indicate that enhanced algorithm has achieved marked improvement metrics such as Precision, Recall, F1, mAP, increases 3.39%, 0.86%, 2.20%, 2.81% respectively, when juxtaposed model. Moreover, external validations, FPS enhancements over YOLOv8, YOLOv5, YOLOX, Faster RCNN, SSD deep-learning models are 6.75 Hz, 10.84 12.79 28.24 21.57 mAP improvements practical 2.79%, 2.92%, 3.08%, 7.07%, 3.84% respectively. refined not only ensures efficient accurate tea-grading recognition but also boasts high rates swift capabilities, establishing foundation development tea-picking robots quality devices.

Language: Английский

SDS-YOLO: An improved vibratory position detection algorithm based on YOLOv11 DOI
Dingran Wang,

Jiasheng Tan,

Hong Wang

et al.

Measurement, Journal Year: 2024, Volume and Issue: unknown, P. 116518 - 116518

Published: Dec. 1, 2024

Language: Английский

Citations

5
LSOD-YOLOv8: Enhancing YOLOv8n with New Detection Head and Lightweight Module for Efficient Cigarette Detection DOI Creative Commons
Yijie Huang, Huimin Ouyang, Xiaodong Miao

et al.

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(7), P. 3961 - 3961

Published: April 3, 2025

Cigarette detection is a crucial component of public safety management. However, detecting such small objects poses significant challenges due to their size and limited feature points. To enhance the accuracy target detection, we propose novel object model, LSOD-YOLOv8 (Lightweight Small Object Detection using YOLOv8). First, introduce lightweight adaptive weight downsampling module in backbone layer YOLOv8 (You Only Look Once version 8), which not only mitigates information loss caused by conventional convolutions but also reduces overall parameter count model. Next, incorporate P2 (Pyramid Pooling Layer 2) neck YOLOv8, blending concepts shared convolutional independent batch normalization design P2-LSCSBD (P2 Layer-Lightweight Shared Convolutional Batch Normalization-based Detection) head. Finally, new function, WIMIoU (Weighted Intersection over Union with Inner, Multi-scale, Proposal-aware Optimization), combining ideas WiseIoU (Wise Union), InnerIoU (Inner MPDIoU (Mean Pairwise Distance resulting improvement without any performance. Our experiments demonstrate that enhances for cigarette specifically.

Language: Английский

Citations

0
Flavor Characteristics of Sun-Dried Green Tea in Different Regions of Yunnan: Metabolite Basis and Soil Influencing Factors DOI Creative Commons
Miao Zhou,

Xiujuan Deng,

Qiaomei Wang

et al.

Foods, Journal Year: 2025, Volume and Issue: 14(7), P. 1280 - 1280

Published: April 7, 2025

To elucidate the regional flavor characteristics of sun-dried green tea (SDT) and their underlying influencing factors, a comprehensive analysis was conducted using metabolomics flavoromics approaches. This study systematically examined SDT samples corresponding garden soils from 13 distinct regions in Yunnan Province. The results revealed that could be classified into two groups based on profiles. Compared to Pa Sha (PS), Bang Dong (BD), Ban Shan (DBS), Guo (DG), Su Hu (SH), Gua Feng Zhai (GFZ), Wu Liang (WLS), Xin Nong (XN), Ba Ka Nuan (BKN), Mang Ang (MA), Man (MN), Bing Dao (BDao), Bin (BS) exhibited significant upregulation polyphenols (TP)/free amino acids (FAA) ratio. former group characterized by sweet mellow taste, while latter displayed stronger taste profile. Furthermore, volatile compounds demonstrated geraniol linalool were significantly upregulated PS, BD, DBS, DG, BS, BDao regions, which associated with tender floral aromas. In contrast, isophorone, 2-pentyl furan, 1-octanol, D-limonene, benzaldehyde markedly enriched XN, BKN, MA, MN, SH, GFZ, WLS contributing honey-like aromatic Altitude mineral element phosphorus are potential key factors affecting differences SDT. Specifically, cultivated at higher altitudes elevated available content greater likelihood accumulating compounds. provides scientific evidence for understanding characteristic profiles across different offering valuable insights differentiation production.

Language: Английский

Citations

0
LCLN-CA: A Survival Regression Analysis-Based Prediction Method for Catechin Content in Yunnan Sun-Dried Tea DOI Creative Commons

Hongxu Li,

Qiaomei Wang,

Houqiao Wang

et al.

Horticulturae, Journal Year: 2024, Volume and Issue: 10(12), P. 1321 - 1321

Published: Dec. 11, 2024

Catechins are pivotal determinants of tea quality, with soil environmental factors playing a crucial role in the synthesis and accumulation these compounds. To investigate impact changes garden environments on catechin content sun-dried tea, this study measured samples corresponding leaves from Nanhua, Yunnan, China. By integrating variations those 17 employing COX regression factor analysis, it was found that pH, organic matter (OM), fluoride, arsenic (As), chromium (Cr) were significantly correlated (p < 0.05). Further, using LASSO for variable selection, model named LCLN-CA constructed four variables including OM, As. The demonstrated high fitting accuracy AUC values 0.674, 0.784, 0.749 intervals CA ≤ 10%, 10% 20%, 20% 30% training set, respectively. validation set showed 0.630, 0.756, 0.723, respectively, indicating well-calibrated curve. Based DynNom framework, visual prediction system Yunnan developed. External test dataset achieved an Accuracy 0.870. This explored relationship between soil-related content, paving new way enhancing practical application value artificial intelligence technology agricultural production.

Language: Английский

Citations

1
Impurity detection of premium green tea based on improved lightweight deep learning model DOI
Zezhong Ding, Mei Wang, Bin Hu

et al.

Food Research International, Journal Year: 2024, Volume and Issue: 200, P. 115516 - 115516

Published: Dec. 15, 2024

Language: Английский

Citations

0
Fresh Tea Leaf-Grading Detection: An Improved YOLOv8 Neural Network Model Utilizing Deep Learning DOI Creative Commons
Zejun Wang, Yuxin Xia,

Houqiao Wang

et al.

Horticulturae, Journal Year: 2024, Volume and Issue: 10(12), P. 1347 - 1347

Published: Dec. 15, 2024

To facilitate the realization of automated tea picking and enhance speed accuracy leaf grading detection, this study proposes an improved YOLOv8 network for fresh recognition. This approach integrates a Hierarchical Vision Transformer using Shifted Windows to replace segments original YOLOv8’s architecture, thereby alleviating computational load dense image processing tasks reducing expenses. The incorporation Efficient Multi-Scale Attention Module with Cross-Spatial Learning serves attenuate influence irrelevant features in complex backgrounds, which turn, elevates model’s detection Precision. Additionally, substitution loss function SIoU facilitates more rapid model convergence precise pinpointing defect locations. empirical findings indicate that enhanced algorithm has achieved marked improvement metrics such as Precision, Recall, F1, mAP, increases 3.39%, 0.86%, 2.20%, 2.81% respectively, when juxtaposed model. Moreover, external validations, FPS enhancements over YOLOv8, YOLOv5, YOLOX, Faster RCNN, SSD deep-learning models are 6.75 Hz, 10.84 12.79 28.24 21.57 mAP improvements practical 2.79%, 2.92%, 3.08%, 7.07%, 3.84% respectively. refined not only ensures efficient accurate tea-grading recognition but also boasts high rates swift capabilities, establishing foundation development tea-picking robots quality devices.

Language: Английский

Citations

0