改进YOLOv8的实时轻量化鲁棒绿篱检测算法

doi:10.19678/j.issn.1000-3428.0069524

摘要/Abstract

摘要：

针对道路两侧绿篱修剪的目标检测过程中对算法实时性、轻量化的要求以及算法在实际检测中的精度和光照鲁棒性问题，提出一种基于YOLOv8n的算法MGW-YOLO，并给出一种新的C2f_ModuGhost+模块来替换主干网络中的C2f模块，其中设计的调制可变形卷积增加了偏移量特征通道数，以加速模型的推理，增强算法实时性。在颈部网络中引入分组空间卷积(GSConv)轻量级卷积技术和slim-neck设计范式，并通过融合标准卷积、深度可分离卷积和Shuffle模块的思想，降低模型的参数量，实现模型的轻量化。设计一种具有双重加权机制的Focal-WIoU损失函数，WIoU中的双层交叉注意力机制可有效降低多个绿篱相连和遮挡时的误检率，并且利用Focal Loss权重因子提升对特殊形状绿篱等难分类样本的检测精度。另外采用TRADES方法的对抗训练策略，在分类问题鲁棒性与精度之间进行有效权衡。实验结果表明，相比基线算法YOLOv8n，MGW-YOLO的mAP@0.5和mAP@0.5 ∶0.95分别提高了3.29和2.87百分点，在无人驾驶底盘上的实验结果表明，MGW-YOLO相较于原始算法的预处理时间、每帧平均推理时间和每帧后处理时间分别降低了0.7 ms、10.7 ms和0.7 ms，检测速度提升了15.7帧/s，适用于绿篱修剪机在道路两侧实时性作业的需求。

关键词: YOLOv8算法, 目标检测, C2f_ModuGhost+模块, 分组空间卷积轻量级卷积, Focal-WIoU损失函数, 对抗训练

Abstract:

This study presents MGW-YOLO, an algorithm based on YOLOv8n. The study aims to address the need for an accurate, real-time, robust, and lightweight algorithm for target detection during hedge trimming on both sides of a road. The study also proposes a new C2f_ModuGhost+ module to replace the C2f module in the backbone network, in which modulated deformable convolution increases the number of offset feature channels, which accelerates model inference and improves the real-time algorithm. The Grouped Spatial Convolution (GSConv) lightweight convolution technique and slim-neck design paradigm are introduced into the neck of the network, which integrates concepts such as standard convolution, depth-separable convolution, and Shuffle module; reduces the number of parameters; and makes the model lightweight. A focal-WIoU loss function with a double weighting mechanism is designed. The two-layer cross-attention mechanism in WIoU effectively reduces the false detection rate when multiple hedges are connected and occluded, and the focal loss weighting factor is utilized to improve the detection accuracy of difficult-to-classify samples such as special-shaped hedges. In addition, the adversarial training strategy of TRADES is adopted to balance robustness and accuracy in the classification problem. Experimental results show that, compared with the baseline algorithm, i.e., YOLOv8n, the mAP@0.5 and mAP@0.5 ∶0.95 of MGW-YOLO increases by 3.29 and 2.87 percentage points, respectively. Experiments on an unmanned chassis show that the pre-processing time, average inference time per frame, and post-processing time per frame of MGW-YOLO are reduced by 0.7 ms, 10.7 ms and 0.7 ms, respectively. The detection speed improves by 15.7 frame/s compared to that of the original algorithm, which is suitable for the real-time operation of hedge trimmers on both sides of a road.

Key words: YOLOv8 algorithm, object detection, C2f_ModuGhost+module, Grouped Spatial Convolution(GSConv)lightweight convolution, Focal-WIoU loss function, adversarial training

张佳承, 韦锦, 陈义时. 改进YOLOv8的实时轻量化鲁棒绿篱检测算法[J]. 计算机工程, 2025, 51(7): 362-374.

ZHANG Jiacheng, WEI Jin, CHEN Yishi. Improved YOLOv8 Real-time Lightweight Robust Hedge Detection Algorithm[J]. Computer Engineering, 2025, 51(7): 362-374.

https://www.ecice06.com/CN/Y2025/V51/I7/362

图/表 19

图1 MGW-YOLO算法网络结构

Fig.1 Network structure of MGW-YOLO algorithm

图2 普通Ghost Bottleneck结构和ModuGhost-bneck结构对比

Fig.2 Comparison between ordinary Ghost Bottleneck structure and ModuGhost-bneck structure

图3 可变形卷积和调制可变形卷积对比

Fig.3 Comparison of deformable convolution and modulated deformable convolution

图4 GSConv结构

Fig.4 GSConv structure

图5 Slim-neck设计范式

Fig.5 Slim-neck design paradigm

图6 自然训练和对抗训练的决策边界对比

Fig.6 Comparison of decision boundaries for natural and adversarial training

图7 绿篱数据集示例

Fig.7 Example of hedge dataset

图8 数据增强操作

Fig.8 Data enhancement operation

图9 YOLOv8n与MGW-YOLO的检测效果对比

Fig.9 Comparison of detection effects between YOLOv8n and MGW-YOLO

图10 无人驾驶电动底盘实验平台

Fig.10 Autonomous driving electric chassis experimental platform

图11 绿篱实时检测效果对比

Fig.11 Comparison of the effects of real-time detection of hedges

参考文献 35

1	陈若虹, 童志成, 吴翠雯. 市政绿篱机械化快速修剪技术分析与应用推广-以西樵镇道路绿化养护项目为例. 现代园艺, 2022(16): 180- 182.
	CHEN R H, TONG Z C, WU C W. Analysis and application promotion of mechanized rapid pruning technology for municipal hedgehogs-taking xiqiao town road greening and maintenance project as an example. Contemporary Horticulture, 2022(16): 180- 182.
2	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector. Berlin, Germany: Springer, 2016.
3	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2017: 2980-2988.
4	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE Press, 2016: 779-788.
5	WANG C Y, BOCHKOVSKIY A, LIAO H M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2023: 7464-7475.
6	REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, USA: IEEE Press, 2017: 7263-7271.
7	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2014: 580-587.
8	GIRSHICK R. Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2015: 1440-1448.
9	REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137- 1149. doi: 10.1109/TPAMI.2016.2577031
10	BRANKO Š, MATAVULJ P, MIMIĆ G, et al. Real-time automatic detection of starch particles in ambient air. Agricultural and Forest Meteorology, 2022, 323, 109034.
11	MEI S, DING W Q, WANG J P. Research on the real-time detection of red fruit based on the you only look once algorithm. Processes, 2024, 12(1): 15.
12	LI S, TAO T, ZHANG Y, et al. YOLO v7-CS: a YOLO v7-based model for lightweight bayberry target detection count. Agronomy, 2023, 13(12): 2952.
13	韦锦, 李正强, 许恩永, 等. 基于DA2-YOLOv4算法绿篱识别研究. 中国农机化学报, 2022, 43(9): 122- 130.
	WEI J, LI Z Q, XU E Y, et al. Research on hedge recognition based on DA2-YOLOv4 algorithm. Journal of Chinese Agricultural Mechanization, 2022, 43(9): 122- 130.
14	JINTASUTTISAK T, EDIRISINGHE E, ELBATTAY A. Deep neural network based date palm tree detection in drone imagery. Computers and Electronics in Agriculture, 2022, 192, 106560.
15	LI R J, LI Y D, QIN W B, et al. Lightweight network for corn leaf disease identification based on improved YOLO v8s. Agriculture, 2024, 14(2): 220.
16	CHEN C P, GUO Z C, ZENG H E, et al. RepGhost: a hardware-efficient ghost module via re-parameterization[EB/OL]. [2024-02-01]. https://arxiv.org/abs/2211.06088v2.
17	LI H L, LI J, WEI H B, et al. Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles[EB/OL]. [2024-02-01]. https://arxiv.org/abs/2206.02424.
18	TONG Z J, CHEN Y H, XU Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[EB/OL]. [2024-02-01]. https://arxiv.org/abs/2301.10051v3.
19	HAN K, WANG Y H, TIAN Q, et al. GhostNet: more features from cheap operations[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2020: 1580-1589.
20	DING X H, GUO Y C, DING G G, et al. ACNet: strengthening the kernel skeletons for powerful CNN via asymmetric convolution blocks[C]//Proceedings of IEEE/CVF International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2019: 1911-1920.
21	DING X H, ZHANG X Y, MA N N, et al. RepVGG: making VGG-style ConvNets great again[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2021: 13733-13742.
22	ZHANG X Y, ZHOU X Y, LIN M X, et al. ShuffleNet: an extremely efficient convolutional neural network for mobile devices[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2018: 6848-6856.
23	MA N N, ZHANG X Y, ZHENG H T, et al. ShuffleNet V2: practical guidelines for efficient CNN architecture design[C]//Proceedings of European Conference on Computer Vision. Berlin, Germany: Springer, 2018: 122-138.
24	MIYATO T, DAI A M, GOODFELLOW I. Adversarial training methods for semi-supervised text classification[EB/OL]. [2024-02-01]. https://arxiv.org/abs/1605.07725v4.
25	GOODFELLOW I J, SHLENS J, SZEGEDY C, et al. Explaining and harnessing adversarial examples[EB/OL]. [2024-02-01]. https://arxiv.org/abs/1412.6572v3.
26	SHAFAHI A, NAJIBI M, GHIASI M A, et al. Adversarial training for free[C]//Proceedings of Advances in Neural Information Processing Systems. Cambridge, USA: MIT Press, 2019: 32-41.
27	ZHU C, CHENG Y, GAN Z, et al. FreeLB: enhanced adversarial training for natural language understanding[EB/OL]. [2024-02-01]. https://arxiv.org/abs/1909.11764v5.
28	LEE S H, BAE S H. AFI-GAN: improving feature interpolation of feature pyramid networks via adversarial training for object detection. Pattern Recognition, 2023, 138, 109365.
29	SHENG J J, ZHANG D W, CHEN J X, et al. Towards universal and sparse adversarial examples for visual object tracking. Applied Soft Computing, 2024, 153, 111252.
30	ZHANG H Y, YU Y D, JIAO J T, et al. Theoretically principled trade-off between robustness and accuracy[EB/OL]. [2024-02-01]. https://arxiv.org/abs/1901.08573v3.
31	CISSE M, BOJANOWSKI P, GRAVE E, et al. Parseval networks: improving robustness to adversarial examples[C]//Proceedings of International Conference on Machine Learning. Washington D. C., USA: IEEE Press, 2017: 854-863.
32	LI B H, HOU Y T, CHE W X. Data augmentation approaches in natural language processing: a survey. AI Open, 2022, 3, 71- 90.
33	LI J F, WEN Y, HE L H. SCConv: spatial and channel reconstruction convolution for feature redundancy[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2023: 6153-6162.
34	KANG M, TING C M, TING F F, et al. RCS-YOLO: a fast andHigh-Accuracy object detector forBrain tumor detection[C]//Proceedings of Conference on Medical Image Computing and Computer Assisted Intervention. Berlin, Germany: Springer, 2023: 600-610.
35	LI C, ZHOU A J, YAO A B. Omni-dimensional dynamic convolution[EB/OL].[2024-02-01]. https://arxiv.org/abs/2209.07947v1.

[1]	宋杰, 徐慧英, 朱信忠, 黄晓, 陈晨, 王泽宇. 基于YOLOv8改进的跌倒检测算法: OEF-YOLO[J]. 计算机工程, 2025, 51(7): 127-139.
[2]	刘旭东, 杨绪兵. L1-OCSVM模型设计及其在林业目标检测中的应用[J]. 计算机工程, 2025, 51(7): 375-384.
[3]	刘春霞, 孟吉星, 潘理虎, 龚大立. 融合RGB与IR图像的遥感小目标检测方法[J]. 计算机工程, 2025, 51(7): 326-338.
[4]	栾孟娜, 郑秋梅, 王风华. 基于DMC-YOLO的交通标志实时检测算法[J]. 计算机工程, 2025, 51(7): 90-99.
[5]	奚琦, 王明杰, 魏敬和, 赵伟. 基于改进YOLOv3的航拍小目标检测算法[J]. 计算机工程, 2025, 51(6): 184-192.
[6]	赵小虎, 谢礼逊, 慕灯聪, 张悦. 基于TCM-YOLO网络的金属表面缺陷检测方法[J]. 计算机工程, 2025, 51(6): 338-348.
[7]	冯晓飞, 谢诚, 张秀振, 董仕奎, 陈军胜, 叶舒, 钟忺. 基于动静结合互学习的预制梁工序检测方法[J]. 计算机工程, 2025, 51(6): 385-394.
[8]	李毅, 徐慧英, 朱信忠, 黄晓, 王舒梦, 李悉钰. 基于YOLOv5n模型改进的口罩检测算法: Mask-YOLO[J]. 计算机工程, 2025, 51(6): 297-310.
[9]	黄琦强, 安国成, 熊刚. 基于视觉-语言预训练模型的开集交通目标检测算法[J]. 计算机工程, 2025, 51(6): 375-384.
[10]	王晓龙, 江波, 罗润书, 安国成. 基于多信息融合的高速收费站拥堵检测算法[J]. 计算机工程, 2025, 51(5): 377-386.
[11]	周思瑜, 徐慧英, 朱信忠, 黄晓, 盛轲, 曹雨淇, 陈晨. 基于改进YOLOv8n的手机屏幕瑕疵检测算法: PGS-YOLO[J]. 计算机工程, 2025, 51(5): 326-339.
[12]	许华杰, 郑力文, 张品, 秦远卓. 基于多维注意力模块的轻量化混凝土裂缝检测方法[J]. 计算机工程, 2025, 51(5): 351-360.
[13]	黄昆, 齐肇建, 王娟敏, 胡倩, 胡伟超, 皮建勇. 基于改进YOLOv8的密集行人检测模型[J]. 计算机工程, 2025, 51(5): 133-142.
[14]	李书玮, 黄正翔, 胡云, 刘兴, 卢笑, 郭畅, 吴成中, 王耀南. 基于无源领域自适应的低光照显著性目标检测[J]. 计算机工程, 2025, 51(4): 75-84.
[15]	王泽宇, 徐慧英, 朱信忠, 黄晓, 梁佳杰, 李琛. 基于改进YOLOv8的轻量化鱼苗检测算法: FD-YOLO[J]. 计算机工程, 2025, 51(4): 327-338.

选择文件类型/文献管理软件名称

选择包含的内容