基于改进YOLOv8s的玫瑰鲜切花分级方法

doi:10.12133/j.smartag.SA202401005

摘要/Abstract

摘要：

[目的/意义] 针对当前玫瑰鲜切花分级仍依赖人工进行简单分级，造成效率低、准确率低等问题，提出一种新的模型Flower-YOLOv8s来实现玫瑰鲜切花的分级检测。 [方法] 以单一背景下单支玫瑰花的花头作为检测目标，将鲜切花分为A、B、C、D四个等级，对YOLOv8s（You Only Look Once version 8 small）模型进行了优化改进。首先，构建了一个全新的玫瑰鲜切花分级检测数据集。其次，在YOLOv8s的骨干网络分别添加CBAM（Convolutional Block Attention Module）和SAM（Spatial Attion Module）两个注意力机制模块进行对比实验；选择SAM模块并对其进一步优化，针对模型轻量化需求，再结合深度可分离卷积模块一起添加到C2f结构中，形成Flower-YOLOv8s模型。 [结果和讨论] 从实验结果来看YOLOv8s添加SAM的模型具有更高的检测精度，mAP@0.5达到86.4%。Flower-YOLOv8s相较于基线模型精确率提高了2.1%，达到97.4%，平均精度均值（mAP）提高了0.7%，同时降低了模型参数和计算量，分别降低2.26 M和4.45 MB；最后使用相同的数据集和预处理方法与Fast-RCNN、Faster-RCNN、SSD、YOLOv3、YOLOv5s和YOLOv8s进行对比实验，证明所提出的实验方法综合强于其他经典YOLO模型。 [结论] 提出的基于改进YOLOv8s的玫瑰鲜切花分级方法研究能有效提升玫瑰鲜切花分级检测的精准度，为玫瑰鲜切花分级检测技术提供一定的参考价值。

关键词: YOLOv8s, 玫瑰鲜切花, 分级检测, 深度学习, SAM, 注意力机制

Abstract:

[Objective] The fresh cut rose industry has shown a positive growth trend in recent years, demonstrating sustained development. Considering the current fresh cut roses grading process relies on simple manual grading, which results in low efficiency and accuracy, a new model named Flower-YOLOv8s was proposed for grading detection of fresh cut roses. [Methods] The flower head of a single rose against a uniform background was selected as the primary detection target. Subsequently, fresh cut roses were categorized into four distinct grades: A, B, C, and D. These grades were determined based on factors such as color, size, and freshness, ensuring a comprehensive and objective grading system. A novel dataset contenting 778 images was specifically tailored for rose fresh-cut flower grading and detection was constructed. This dataset served as the foundation for our subsequent experiments and analysis. To further enhance the performance of the YOLOv8s model, two cutting-edge attention convolutional block attention module (CBAM) and spatial attention module (SAM) were introduced separately for comparison experiments. These modules were seamlessly integrated into the backbone network of the YOLOv8s model to enhance its ability to focus on salient features and suppressing irrelevant information. Moreover, selecting and optimizing the SAM module by reducing the number of convolution kernels, incorporating a depth-separable convolution module and reducing the number of input channels to improve the module's efficiency and contribute to reducing the overall computational complexity of the model. The convolution layer (Conv) in the C2f module was replaced by the depth separable convolution (DWConv), and then combined with Optimized-SAM was introduced into the C2f structure, giving birth to the Flower-YOLOv8s model. Precision, recall and F₁ score were used as evaluation indicators. [Results and Discussions] Ablation results showed that the Flower-YOLOv8s model proposed in this study, namely YOLOv8s+DWConv+Optimized-SAM, the recall rate was 95.4%, which was 3.8% higher and the average accuracy, 0.2% higher than that of YOLOv8s with DWConv alone. When compared to the baseline model YOLOv8s, the Flower-YOLOv8s model exhibited a remarkable 2.1% increase in accuracy, reaching a peak of 97.4%. Furthermore, mAP was augmented by 0.7%, demonstrating the model's superior performance across various evaluation metrics. The effectiveness of adding Optimized-SAM was proved. From the overall experimental results, the number of parameters of Flower-YOLOv8s was reduced by 2.26 M compared with the baseline model YOLOv8s, and the reasoning time was also reduced from 15.6 to 5.7 ms. Therefore, the Flower-YOLOv8s model was superior to the baseline model in terms of accuracy rate, average accuracy, number of parameters, detection time and model size. The performances of Flower-YOLOv8s network were compared with other target detection algorithms of Fast-RCNN, Faster-RCNN and first-stage target detection models of SSD, YOLOv3, YOLOv5s and YOLOv8s to verify the superiority under the same condition and the same data set. The average precision values of the Flower-YOLOv8s model proposed in this study were 2.6%, 19.4%, 6.5%, 1.7%, 1.9% and 0.7% higher than those of Fast-RCNN, Faster-RCNN, SSD, YOLOv3, YOLOv5s and YOLOv8s, respectively. Compared with YOLOv8s with higher recall rate, Flower-YOLOv8s reduced model size, inference time and parameter number by 4.5 MB, 9.9 ms and 2.26 M, respectively. Notably, the Flower-YOLOv8s model achieved these improvements while simultaneously reducing model parameters and computational complexity. [Conclusions] The Flower-YOLOv8s model not only demonstrated superior detection accuracy but also exhibited a reduction in model parameters and computational complexity. This lightweight yet powerful model is highly suitable for real-time applications, making it a promising candidate for flower grading and detection tasks in the agricultural and horticultural industries.

Key words: YOLOv8s, fresh cut roses, hierarchical detection, deep learning, SAM, attention mechanism

张玉玉, 邴树营, 纪元浩, 严蓓蓓, 许金普. 基于改进YOLOv8s的玫瑰鲜切花分级方法[J]. 智慧农业(中英文), 2024, 6(2): 118-127.

ZHANG Yuyu, BING Shuying, JI Yuanhao, YAN Beibei, XU Jinpu. Grading Method of Fresh Cut Rose Flowers Based on Improved YOLOv8s[J]. Smart Agriculture, 2024, 6(2): 118-127.

图/表 13

表1

图1

图2

表2

表3

图3

图4

图5

图6

图7

表4

表5

图8

参考文献 25

1	中国报告大厅. 2022年玫瑰行业分析[EB/OL]. [2023-12-02].
2	光明网. 年产量达180亿枝:探索云南鲜切花产业背后的科技力量[EB/OL]. (2023-08-22).
3	DENG L M, LI J, HAN Z Z. Online defect detection and automatic grading of carrots using computer vision combined with deep learning methods[J]. LWT-food science and technology, 2021, 149: ID 111832.
4	STEINMETZ V, DELWICHE M J, GILES D K, et al. Sorting cut roses with machine vision[J]. Transactions of the ASAE, 1994, 37(4): 1347-1353.
5	HIARY H, SAADEH H, SAADEH M, et al. Flower classification using deep convolutional neural networks[J]. IET computer vision, 2018, 12(6): 855-862.
6	METE B R, ENSARI T. Flower classification with deep CNN and machine learning algorithms[C]// 2019 3rd International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT). Piscataway, New Jersey, USA: IEEE, 2019: 1-5.
7	赵晓龙. 基于神经网络的玫瑰花图像等级分类识别研究[D]. 昆明: 云南大学, 2021.
	ZHAO X L. Research on classification and recognition of rose image based on neural network[D]. Kunming: Yunnan University, 2021.
8	孙鑫岩. 基于深度学习的鲜切花分级算法[D]. 南京: 南京林业大学, 2022.
	SUN X Y. Fresh cut flower grading algorithm based on deep learning[D]. Nanjing: Nanjing Forestry University, 2022.
9	吴宇. 基于深度学习的玫瑰鲜切花分级研究[D]. 昆明: 云南大学, 2021.
	WU Y. Quality Grading Evaluation of Cut Rose based on Deep Learning[D]. Kunming: Yunnan University, 2021.
10	JOCHER G, CHAURASIA A, QIU J. YOLO by Ultralytics[EB/OL]. (2023-06-30) [2023-12-31].
11	国家质量技术监督局. 主要花卉产品等级第1部分:鲜切花: GB/T 18247.1—2000 [S]. 北京: 中国标准出版社, 2001.
	State Bureau of Quality and Technical Supervision of the People's Republic of China. Product grade for major ornamental plants-Part 1: Cut flowers: GB/T 18247.1—2000 [S]. Beijing: Standards Press of China, 2001.
12	FU J, LIU J, TIAN H J, et al. Dual attention network for scene segmentation[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, New Jersey, USA: IEEE, 2019: 3141-3149.
13	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]// European Conference on Computer Vision. Berlin, German: Springer, 2018: 3-19.
14	SIFRE L. Rigid-motion scattering for image classification[D]. Paris: École Polytechnique, 2014.
15	REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, New Jersey, USA: IEEE, 2017: 6517-6525.
16	JIANG Z, ZHAO L, LI S, et al. Real-time object detection method based on improved YOLOv4-tiny[EB/OL]. arXiv: 2011.04244, 2020.
17	GLENN J.YOLOv 5. [EB/OL]. Git Code,2020. [2023-12-31]
18	LIU S, QI L, QIN H F, et al. Path aggregation network for instance segmentation[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, New Jersey, USA: IEEE, 2018: 8759-8768.
19	REN S, HE K, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[C]// Advances in Neural Information Processing Systems. Cambridge, USA: MIT Press, 2015: 91-99.
20	LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C]// European Conference on Computer Vision. Berlin, German: Springer, 2016: 21-37.
21	REDMON J, FARHADI A. YOLOv3: An incremental improvement[EB/OL]. arXiv: 1804.02767, 2018.
22	高芳芳, 武振超, 索睿, 等. 基于深度学习与目标跟踪的苹果检测与视频计数方法[J]. 农业工程学报, 2021, 37(21): 217-224.
	GAO F F, WU Z C, SUO R, et al. Apple detection and counting using real-time video based on deep learning and object tracking[J]. Transactions of the Chinese society of agricultural engineering, 2021, 37(21): 217-224.
23	施行. 基于视觉技术的红提串品质无损检测与分级[D]. 武汉: 华中农业大学, 2021.
	SHI H. Non-destructive detection and grading of the quality of red grape strings based on visual technology[D]. Wuhan: Huazhong Agricultural University, 2021.
24	王俊杰. 基于深度学习的红枣缺陷检测分级技术研究[D]. 西安: 陕西科技大学, 2020.
	WANG J J. Research on defect detection and classification of jujube based on deep learning[D]. Xi'an: Shaanxi University of Science & Technology, 2020.
25	AMAN M. Postharvest loss estimation of cut rose (Rosa hybrida) flower farms: Economic analysis in East Shoa Zone, Ethiopia[J]. International journal of sustainable economy, 2014, 6(1): ID 82.

等级	分级标准
A级	花头大，花色花型正常；无弯头，开放度基本一致，基本无色差；叶片无损伤，无虫害，成熟度达到季节采收要求；茎秆粗细均匀，叶面清秀无污染
B级	花头大，花色花型正常，有轻微损伤；无弯头，开放度大部分一致，基本无色差；叶片基本无损伤，无虫害，成熟度达到季节采收要求；茎秆粗细均匀，叶面清秀无污染
C级	花色、花型有偏差，可能有双心花，花头、花径、叶有轻微损伤；轻微病虫害、弯头、缺陷，开放度大部分一致；部分茎秆弯曲，粗细不均匀，叶面有轻微药斑、病斑
D级	花色、花型有畸形或擦痕，花头、花径、叶片有损伤；大部分为侧枝切花，有病虫害、弯头、缺陷，开放度一般；茎秆多数细短弯曲，叶面有药斑病斑

过程	A级/张	B级/张	C级/张	D级/张	合计/张
增强前	46	95	107	76	324
增强后	114	231	244	189	778

参数	数值
学习率（Learning Rate）	0.000 1
图像大小（Image Size）/像素	640×640
梯度动量（Momentum）	0.937
Batch Size	16
迭代轮数（Epochs）	200
优化器（Optimizer）	Adam

模型名称	精确率/%	召回率/%	mAP@0.5/%	参数量/M	推理时间/ms	模型大小/MB
YOLOv8s	95.3	96.5	82.4	11.13	15.6	22.5
YOLOv8s+DWConv	97.2	91.6	82.9	11.13	9.2	22.6
YOLOv8s+DWConv+Optimized-SAM（Flower-YOLOv8s）	97.4	95.4	83.1	8.87	5.7	18.0

模型名称	准确率/%	召回率/%	mAP@0.5/%	mAP@0.5∶0.95/%	F ₁值/%	参数量/10⁶	推理时间/ms	模型大小/MB
Fast-RCNN	86.8	81.4	80.5	76.1	0.55	193.9	36.5	102.35
Faster-RCNN	93.3	87.2	63.7	58.9	0.49	254.6	48.8	142.08
SSD	95.8	80.6	76.3	68.8	0.66	106.4	25.3	97.42
YOLOv3	93.6	76.9	81.4	71.8	0.74	207.8	22.6	103.67
YOLOv5s	86.1	74.4	81.2	71.0	0.75	22.2	5.9	10.92
YOLOv8s	95.3	96.5	82.4	72.2	0.75	22.5	15.6	11.13
Flower-YOLOv8s	97.4	95.4	83.1	72.5	0.77	18.0	5.7	8.87