基于YOLOv8-MRF的黄瓜生育期检测方法

Mobile device-based cucumber growth stage detection method based on YOLOv8-MRF

摘要

针对温室环境下黄瓜生育期检测精度低、边缘设备难以兼顾检测精度与实时性等问题,本文提出一种基于YOLOv8-MRF的黄瓜生育期检测方法。首先,在骨干网络中引入了混合局部通道注意力机制(mixed local channel attention,MLCA),以增强模型的特征信息提取能力;其次,采用感受野坐标注意力卷积(Receptive field coordinates attention convolution,RFCAConv)融合多尺度上下文信息,以增强模型多尺度信息感知能力;随后,更换损失函数为Focal-EIoU,增强困难样本的权重贡献,以提升模型收敛速度与检测准确度;最后,采用模型剪枝与模型量化技术降低模型计算量,并将YOLOv8-MRF模型部署至RK3588开发板。试验结果表明:YOLOv8-MRF模型的准确率、召回率、平均精度均值与F1分数分别为87.0%、83.8%、88.7%、85.3%;经LAMP剪枝与量化,模型参数量降至1.12 M,权重仅2.6 MB,减小了边缘设备的运算压力,模型部署至RK3588边缘设备后,平均精度均值达到88.4%,平均帧率为35 f/s,相较于YOLOv8n模型,其差值为2.5%和20 f/s,可实现黄瓜生育期的实时、精准检测,为设施农业黄瓜生育期模型的边缘部署与工程化应用提供了技术支撑。

Abstract

To address the issues of low detection accuracy for cucumber growth stages in greenhouse environments and the difficulty for edge devices to balance detection accuracy and real-time performance, this study proposes a cucumber growth stage detection method based on YOLOv8-MRF. Firstly, a mixed local channel attention mechanism (MLCA) is integrated into the backbone network to enhance the model's feature information extraction capability; secondly, receptive field coordinates attention convolution (RFCAConv) is used to fuse multi-scale contextual information, thereby enhancing the model's multi-scale information perception ability. Subsequently, the loss function is replaced with Focal-EIoU to enhance the weight contribution of hard samples, thereby improving the model's convergence speed and detection accuracy. Finally, model pruning and model quantization techniques are adopted to reduce the model's computational load, and the YOLOv8-MRF model is deployed onto an RK3588 development board. The experimental results show that the accuracy, recall rate, mean average precision and F1 score of the YOLOv8-MRF model are 87.0%, 83.8%, 88.7% and 85.3%, respectively. After LAMP pruning and quantization, the model's parameter count is reduced to 1.12 M with a weights of only 2.6 MB, reducing the computational pressure on edge devices. After the model is deployed onto the RK3588 edge device, the mean average precision reaches 88.4%, and the average frame rate is 35 f/s. Compared with the YOLOv8n model, the differences are 2.5% and 20 f/s, respectively. It enables real-time and accurate detection of cucumber growth stages, providing technical support for the edge deployment and engineering application of cucumber growth stage models in facility agriculture.

关键词

生育期识别 / 边缘设备 / LAMP剪枝 / 模型量化

Key words

growth stage recognition / edge devices / LAMP pruning / model quantization

[1] 周海霞,杨凡,张果,等.河南省设施黄瓜产业现状、存在问题与高质量发展对策[J].中国瓜菜,2025,38(4):198-205.
Zhou Haixia, Yang Fan, Zhang Guo, et al. The current situation, existing problems, and countermeasures for high quality development of facility cucumber industry in Henan province[J]. China Cucurbits and Vegetables, 2025, 38(4): 198-205.
[2] Hossain S A A M, Wang L X, Chen T T, et al. Leaf area index assessment for tomato and cucumber growing period under different water treatments[J]. Plant, Soil and Environment, 2017, 63: 461-467.
[3] 董适,赵国瑞,苟豪,等.基于改进RT-Detr的黄瓜果实选择性采摘识别方法[J].农业工程学报,2025,41(1):212-220.
Dong Shi, Zhao Guorui, Gou Hao, et al. Identifying cucu-mber fruits during selective picking using improved RTDetr[J]. Transactions of the Chinese Society of Agricultural Engineering, 2025,41(1):212-220.
[4] Tan K Z, Lee W S, Gan H, et al. Recognising blueberry fruit of different maturity using histogram oriented gradients and colour features in outdoor scenes[J]. Biosystems Engineering, 2018, 176: 59-72.
[5] 夏雪,周国民,丘耘,等.自然环境下果实作业机器人幼果期苹果侦测方法[J].中国农业科技导报,2018,20(5):64-74.
Xia Xue, Zhou Guomin, Qiu Yun, et al. Detection of young green apples for fruit robot in natural scene[J]. Journal of Agricultural Science and Technology,2018,20(5):64-74.
[6] 杨林,曾大鑫,边银丙,等.改进YOLOv5的香菇子实体生育期识别方法[J].农业工程学报,2024,40(9):182-189.
Yang Lin, Zeng Daxin, Bian Yinbing, et al. Recognizing fruiting body growth period of Lentinus edodes using improved YOLOv5[J]. Transactions of the Chinese Society of Agricultural Engineering, 2024,40(9):182-189.
[7] 黄家才,赵雪迪,高芳征,等.基于改进YOLOv5s的草莓多阶段识别检测轻量化算法[J].农业工程学报,2023,39(21):181-187.
Huang Jiacai, Zhao Xuedi, Gao Fangzheng, et al. Recognizing and detecting the strawberry at multi-stages using improved lightweight YOLOv5s[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023,39(21):181-187.
[8] Bai Y H, Guo Y X, Zhang Q, et al. Multi-network fusion algorithm with transfer learning for green cucumber segmentation and recognition under complex natural environment[J]. Computers and Electronics in Agriculture, 2022, 194: 106789.
[9] 高文瑞,徐刚,孙艳军,等.设施黄瓜果实发育过程中品质及形态变化规律的研究[J].金陵科技学院学报,2011,27(1):44-47.
Gao Wenrui, Xu Gang, Sun Yanjun, et al. The Fruit quality and morphological changes during the development process of cucumber fruits[J]. Journal of Jinling Institute of Technology, 2011,27(1):44-47.
[10] 杨悦.黄瓜果实发育过程纵横径变化规律及其器官形状相关性研究[D].杨凌:西北农林科技大学,2018.
[11] Wang Q L, Wu B G, Zhu P F, et al. ECA-Net: Efficient channel attention for deep convolutional neural networks[J]. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2020: 11531-11539.
[12] Hu J, Shen L, Sun G, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023.
[13] Wan D H, Lu R S, Shen S Y, et al. Mixed local channel attention for object detection[J]. Engineering Applications of Artificial Intelligence, 2023, 123(PC): 106442.
[14] Zhang X, Liu C, Yang D G, et al. RFAConv: innovating spatial attention and standard convolutional operation[EB/OL].2023-04-06[2025-06-25].https://arxiv.org/abs/2304.03198.
[15] Hou Q B, Zhou D Q, Feng J S, et al. Coordinate attention for efficient mobile network design[EB/OL].2021-03-04[2025-06-25].https://arxiv.org/abs/2103.02907.
[16] Zheng Z H, Wang P, Liu W, et al. Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression[J]. Proceedings of the Aaai Conference on Artificial Intelligence, 2020, 34(7): 2993-13000.
[17] Zhang Y F, Ren W Q, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157.
[18] 刘诗怡,胡滨,赵春.基于改进YOLOv7的黄瓜叶片病虫害检测与识别[J].农业工程学报,2023,39(15):163-171.
Liu Shiyi, Hu Bin, Zhao Chun, et al. Detection and identi-fication of cucumber leaf diseases based improved YOLOv7[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023,39(15):163-171.
[19] Lee J, Park S, Mo S, et al. Layer adaptive sparsity for the magnitude based Pruning[EB/OL].arXiv,2020[2025-12-08].https://arxiv.org/abs/2010.07611.
[20] 施国英,纪嘉鹏,李天华,等.基于改进YOLOv8s的杭白菊检测与花期分类[J].农业工程学报,2025,41(7):192-199.
Shi Guoying, Ji Jiapeng, Li Tianhua, et al. Detecting chrysanthemum to classify flowering stages using improved YOLOv8s[J]. Transactions of the Chinese Society of Agricultural Engineering, 2025, 41(7): 192-199.

基金资助

山东省重大科技创新工程项目(2022CXGC020708)