[1]宋现雪,郑崇珂,邢凡彬,等.基于改进YOLO11n的大田稻穗检测与计数[J].江苏农业学报,2026,42(07):1408-1418.[doi:doi:10.3969/j.issn.1000-4440.2026.07.012]
　SONG Xianxue,ZHENG Chongke,XING Fanbin,et al.Field rice panicle detection and counting based on improved YOLO11n[J].,2026,42(07):1408-1418.[doi:doi:10.3969/j.issn.1000-4440.2026.07.012]
点击复制

基于改进YOLO11n的大田稻穗检测与计数()

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
42
期数:
2026年07期
页码:
1408-1418
栏目:
农业信息工程
出版日期:
2026-07-31

文章信息/Info

Title:
Field rice panicle detection and counting based on improved YOLO11n
作者:
宋现雪1; 2; 郑崇珂3; 邢凡彬1; 2; 王鑫怡1; 2; 谢先芝3; 郑纪业1; 2; 郑世玲1; 张霞1
(1.聊城大学物理科学与信息工程学院,山东聊城252000；2.山东省农业科学院农业信息与经济研究所,山东济南250100；3.山东省农业科学院湿地农业与生态研究所,山东济南250100)
Author(s):
SONG Xianxue1; 2; ZHENG Chongke3; XING Fanbin1; 2; WANG Xinyi1; 2; XIE Xianzhi3; ZHENG Jiye1; 2; ZHENG Shiling1; ZHANG Xia1
(1.School of Physical Science and Information Engineering, Liaocheng University, Liaocheng 252000, China;2.Institute of Agricultural Information and Economics, Shandong Academy of Agricultural Sciences, Jinan 250100, China;3.Institute of Wetland Agriculture and Ecology, Shandong Academy of Agricultural Sciences, Jinan 250100, China)
关键词:
稻穗计数; 轻量化目标检测; 注意力机制; YOLO
Keywords:
rice panicle counting; lightweight object detection; attention mechanism; YOLO
分类号:
S126
DOI:
doi:10.3969/j.issn.1000-4440.2026.07.012
文献标志码:
A
摘要:
水稻稻穗的检测与计数对产量估算和品种选育具有重要意义。大田环境下植株密集、稻穗尺寸小,且常因与叶片交错重叠遮挡严重,因此在复杂场景中实现稻穗的高精度识别与自动计数面临挑战。为此,本研究提出一种融合轻量化检测头与高效注意力机制的检测模型WEL-YOLO,对原模型YOLO11n进行网络结构优化、损失函数改进和模型轻量化。首先,为提高模型对稻穗的特征提取能力,在主干部分C3k2模块中引入高效多尺度注意力(EMA)机制。其次,更换损失函数为智能交并比-归一化高斯瓦瑟斯坦距离(WIoU-NWD),提高定位精度的同时,解决小目标漏检、误检问题。最后,引入轻量级可伸缩共享卷积检测头(LSCD),在保证模型精度的同时进一步降低模型复杂度。在本研究所采用的数据集上进行验证,结果表明WEL-YOLO模型的精确率、召回率、平均检测精度分别为87.1%、81.6%、90.0%；与基准模型相比,精确率和平均检测精度分别提升1.6个百分点和0.9个百分点；模型参数量、计算复杂度和模型权重分别降低6.2%、11.1%、7.3%；均方根误差、平均绝对误差、计数准确率分别为5.59、4.15、80.17%。结果表明,WEL-YOLO模型在提升稻穗检测性能的同时显著降低资源消耗,为水稻稻穗计数的轻量化提供了参考。
Abstract:
The detection and counting of rice panicles are crucial for yield estimation and variety breeding. In field conditions, dense plant growth, small panicle sizes, and severe overlapping with leaves create challenges for high-precision recognition and automatic counting in complex scenarios. To address this, this study proposed a detection model WEL-YOLO, integrating a lightweight detection head and an efficient attention mechanism. The model optimized the network architecture, improved the loss function, and reduced computational complexity compared to the original YOLO11n. First, the C3k2 module in the main trunk introduced an EMA (Efficient multi-scale attention) mechanism to enhance panicle feature extraction. Second, the loss function was replaced with WIoU-NWD (Wise intersection over union-normalized Gaussian Wasserstein distance), which improved localization accuracy while addressing the missed detection and false detection of small objects. Finally, the introduced lightweight scalable shared convolutional detection head (LSCD) further reduced model complexity without compromising accuracy. In the dataset analyzed, the WEL-YOLO model achieved 87.1% accuracy, 81.6% recall, and 90.0% average detection precision, representing 1.6 and 0.9 percentage points improvement in accuracy and average detection precision over the baseline model, respectively. The model demonstrated 6.2% reduction in parameter quantity, 11.1% decrease in computational complexity, and 7.3% reduction in weight values. Its performance metrics included 5.59 root mean square error, 4.15 mean absolute error, and 80.17% count accuracy. These results demonstrate that WEL-YOLO enhances rice panicle detection performance while significantly reducing resource consumption, providing valuable insights for lightweight and practical applications in rice panicle counting.

参考文献/References:

[1]刘泽钰,周云成,梁铖玮,等. 基于可变形卷积的稻粒在穗计数方法[J]. 农业机械学报,2025,56(3):363-373.
[2]Wang G, Chen Y F, An P, et al. UAV-YOLOv8:a small-object-detection model based on improved YOLOv8 for UAV aerial photography scenarios[J]. Sensors,2023,23(16):7190.
[3]Qu F H, Li H L, Wang P, et al. Rice spike identification and number prediction in different periods based on UAV imagery and improved YOLOv8[J]. Computers,Materials & Continua,2025,84(2):3911-3925.
[4]Chen Y, Xin R, Jiang H Y, et al. Refined feature fusion for in-field high-density and multi-scale rice panicle counting in UAV images[J]. Computers and Electronics in Agriculture,2023,211:108032.
[5]朱家微,江朝晖,洪石兰,等. 基于神经架构搜索的灌浆期水稻稻穗分割及特征分析[J]. 激光与光电子学进展,2022,59(22):182-188.
[6]欧阳春凡,高嘉正,陈桥,等. 大视场下火龙果目标检测与计数方法[J]. 中国农业科技导报(中英文),2025,27(8):100-109.
[7]Shao H M, Tang R, Lei Y J, et al. Rice ear counting based on image segmentation and establishment of a dataset[J]. Plants,2021,10(8):1625.
[8]赵静,李京谦,杨蕾,等. 基于PConv-CGLU与重参数检测头的轻量化膜下棉苗实时检测算法[J]. 农业工程学报, 2025,41(18)：151-162.
[9]刘鑫,马本学,李玉洁,等. 基于改进YOLOv7-ByteTrack的干制哈密大枣缺陷检测与计数系统[J]. 农业工程学报,2024,40(3):303-312.
[10]鲍文霞,谢文杰,胡根生,等. 基于TPH-YOLO的无人机图像麦穗计数方法[J]. 农业工程学报,2023,39(1):155-161.
[11]王宏乐,叶全洲,王兴林,等. 基于YOLOv7的无人机影像稻穗计数方法研究[J]. 广东农业科学,2023,50(7):74-82.
[12]Ramachandran A, Sendhil Kumar K S. Border sensitive knowledge distillation for rice panicle detection in UAV images[J]. Computers,Materials & Continua,2024,81(1):827-842.
[13]张远琴,肖德琴,陈焕坤,等. 基于改进Faster R-CNN的水稻稻穗检测方法[J]. 农业机械学报,2021,52(8):231-240.
[14]蔡竹轩,蔡雨霖,曾凡国,等. 基于改进YOLOv51的田间水稻稻穗识别[J]. 华南农业大学学报,2024,45(1):108-115.
[15]肖伸平,赵倩颖,曾甲元,等. 基于YOLO-DCL的复杂环境油茶果遮挡检测与计数研究[J]. 农业机械学报,2024,55(10):318-326,480.
[16]姜海燕,徐灿,陈尧,等. 基于田间图像的局部遮挡小尺寸稻穗检测和计数方法[J]. 农业机械学报,2020,51(9):152-162.
[17]Wei L L, Luo Y S, Xu L Z, et al. Deep convolutional neural network for rice density prescription map at ripening stage using unmanned aerial vehicle-based remotely sensed images[J]. Remote Sensing,2021,14(1):46.
[18]马小林,王梦麟,旷海兰,等. 基于YOLOv8n改进的蚕虫检测与计数方法[J]. 农业工程学报,2024,40(15):143-151.
[19]翟肇裕,张梓涵,徐焕良,等. YOLO算法在动植物表型研究中应用综述[J]. 农业机械学报,2024,55(11):1-20.
[20]杨万里,段凌凤,杨万能. 基于深度学习的水稻表型特征提取和穗质量预测研究[J]. 华中农业大学学报,2021,40(1):227-235.
[21]Madec S, Jin X L, Lu H, et al. Ear density estimation from high resolution RGB imagery using deep learning technique[J]. Agricultural and Forest Meteorology,2019,264:225-234.
[22]Khanam R, Hussain M. YOLOv11:an overview of the key architectural enhancements [EB/OL]. arXiv:2410.17725. (2024-10-23)
[2025-05-26]. https://doi.org/10.48550/arXiv.2410.17725.
[23]王如梦,王赫,武国庆,等. 基于WTEMA-YOLOv11的房屋裂缝检测方法[J]. 电子制作,2025,33(15):33-38.
[24]Zheng Z H, Zhao J G, Fan J J. YOLO-GML:an object edge enhancement detection model for UAV aerial images in complex environments[J]. PLoS One,2025,20(7):e0328070.
[25]Hu D A, Yu M, Wu X Y, et al. DGW-YOLOv8:a small insulator target detection algorithm based on deformable attention backbone and WIoU loss function[J]. IET Image Processing,2024,18(4):1096-1108.
[26]严嘉旭,苏天康,宋慧慧. 改进YOLOv8的轻量化无人机航拍目标检测[J]. 计算机系统应用,2025,34(9):151-161.

备注/Memo

备注/Memo:
收稿日期：2025-10-27基金项目：国家重点研发计划项目(2024YFF1000401)作者简介：宋现雪(2002-),男,山东菏泽人,硕士研究生,主要研究方向为智慧农业。(E-mail)15264027846@163.com通讯作者：谢先芝,(E-mail)xzhxie2010@163.com；郑纪业,(E-mail)jiyezheng@163.com；郑世玲,(E-mail)zhengshiling@lcu.edu.cn
更新日期/Last Update: 2026-08-21