融合多尺度特征与注意力机制的改进YOLOv12农作物虫害检测方法

doi:10.14088/j.cnki.issn0439-8114.2026.03.030

湖北农业科学 ›› 2026, Vol. 65 ›› Issue (3): 190-196.doi: 10.14088/j.cnki.issn0439-8114.2026.03.030

融合多尺度特征与注意力机制的改进YOLOv12农作物虫害检测方法

于丽敏^1,2, 吴奇峰², 曲荣芳¹, 白如霄², 李全胜², 安晓飞³, 张领先⁴

1.山东农业工程学院,济南 250100;
2.新疆农垦科学院,新疆石河子 832000;
3.国家农业智能装备工程技术研究中心,北京 100097;
4.中国农业大学,北京 100083

收稿日期:2025-12-01 出版日期:2026-03-25 发布日期:2026-04-09
通讯作者: 吴奇峰（1978-）,男,新疆石河子人,副研究员,博士,主要从事农业工程研究,（电话）18094801555（电子信箱）1548611544@qq.com。
作者简介:于丽敏（1978-）,女,山东济南人,副教授,博士,主要从事智慧农业研究,（电话）18615177887（电子信箱）18615177887@163.com。
基金资助:
山东大学访问学者项目（202306）; 山东农业工程学院青年教师科研项目（QNKJZ202305）; 新疆BT农业科技创新工程专项项目（NCG202507）

Improved YOLOv12 with multi-scale feature fusion and attention mechanism for crop pest detection

YU Li-min^1,2, WU Qi-feng², QU Rong-fang¹, BAI Ru-xiao², LI Quan-sheng², AN Xiao-fei³, ZHANG Lin-xian⁴

1. Shandong Agriculture and Engineering University, Jinan 250100, China;
2. Xinjiang Academy of Agricultural and Reclamation Sciences, Shihezi 832000, Xinjiang, China;
3. National Engineering Research Center of Intelligent Equipment for Agriculture, Beijing 100097, China;
4. China Agricultural University, Beijing 100083, China

Received:2025-12-01 Published:2026-03-25 Online:2026-04-09

摘要/Abstract

摘要： 为解决农作物虫害目标小、类别多、光照变化大和背景干扰强等问题,基于YOLOv12提出一种面向复杂农田场景的农作物虫害目标检测方法CPD-YOLO（Crop Pest Detection YOLO）。该方法在不显著增加计算开销的前提下,通过构建轻量化高分辨率特征分支、引入轻量化BiFPN多尺度特征融合模块,并设计面向小目标的Anchor-Free检测头,以提升细粒度特征保留、多尺度信息交互与小目标定位回归能力。试验在涵盖水稻、玉米、豆类等102类主要农作物虫害的数据集上进行,结果表明,相较基线YOLOv12,CPD-YOLO在mAP@0.5上提升约3.6个百分点,在mAP@0.5∶0.95上提升约4.2个百分点。与SSD、Faster R-CNN、YOLOv5、YOLOv7、YOLOv8以及YOLOv9相比,CPD-YOLO在mAP@0.5上分别提高约15.3、9.6、7.9、5.1、5.4和4.2个百分点,同时保持接近基线YOLOv12的实时推理速度。研究提出的CPD-YOLO能够有效改善复杂农田条件下的虫害识别性能,可为农作物虫害智能监测和精准防控提供可靠技术支撑。

关键词: 虫害检测, YOLOv12, 小目标检测, 多尺度特征融合, 深度学习

Abstract: To address the challenges of small pest targets, numerous categories, large illumination variations, and strong background interference in complex farmland environments,the CPD-YOLO (Crop Pest Detection YOLO), an improved crop pest detection method based on YOLOv12, was proposed. Without a notable increase in computational cost, CPD-YOLO incorporated a lightweight high-resolution feature branch, a lightweight BiFPN-based multi-scale feature fusion module, and an anchor-free detection head tailored for small objects to enhance fine-grained feature preservation, multi-scale contextual interaction, and small target localization regression ability. Experiments were conducted on a dataset covering 102 major crop pest categories, including pests of rice, maize, and legumes. The results showed that, compared with the YOLOv12 baseline, CPD-YOLO improved mAP@0.5 by approximately 3.6 percentage points and mAP@0.5∶0.95 by approximately 4.2 percentage points. Compared with SSD,Faster R-CNN,YOLOv5,YOLOv7,YOLOv8 and YOLOv9, CPD-YOLO further increased mAP@0.5 by about 15.3,9.6,7.9,5.1,5.4 and 4.2 percentage points, respectively, while maintaining real-time inference speed close to the YOLOv12 baseline. These results demonstrated that CPD-YOLO could effectively improve pest detection performance under complex field conditions and provided reliable technical support for intelligent monitoring and precision control of crop pest.

Key words: pest detection, YOLOv12, small object detection, multi-scale feature fusion, deep learning

中图分类号:

S435

于丽敏, 吴奇峰, 曲荣芳, 白如霄, 李全胜, 安晓飞, 张领先. 融合多尺度特征与注意力机制的改进YOLOv12农作物虫害检测方法[J]. 湖北农业科学, 2026, 65(3): 190-196.

YU Li-min, WU Qi-feng, QU Rong-fang, BAI Ru-xiao, LI Quan-sheng, AN Xiao-fei, ZHANG Lin-xian. Improved YOLOv12 with multi-scale feature fusion and attention mechanism for crop pest detection[J]. HUBEI AGRICULTURAL SCIENCES, 2026, 65(3): 190-196.

参考文献

[1] WANG S H, XU D C, LIANG H J, et al.Advances in deep learning applications for plant disease and pest detection: A review[J]. Remote sensing, 2025, 17(4): 698.
[2] MA C S, WANG B X, WANG X J,et al.Crop pest responses to global changes in climate and land management[J]. Nature reviews earth and environment, 2025, 6(4): 264-283.
[3] KIOBIA D O, MWITTA C J, FUE K G,et al.A review of successes and impeding challenges of IoT-based insect pest detection systems for estimating agroecosystem health and productivity of cotton[J]. Sensors, 2023, 23(8): 4127.
[4] 陈泓润, 朱洪前, 林洁如, 等. 基于YOLOv11-RFDY的苹果病虫害小目标检测[J/OL]. 林业工程学报, 1-9[2026-01-06].https://doi.org/10.13360/j.issn.2096-1359.202412015.
[5] 石成琪, 胡亚琦, 张冰涛. VitiDetect-net:葡萄叶片病虫害的高精度检测算法[J/OL]. 计算机科学与探索, 1-16[2026-01-06].https://link.cnki.net/urlid/11.5602.tp.20250926.1632.005.
[6] 熊彬, 魏登峰, 张建民. 基于改进YOLOv8n的轻量化棉花病虫害检测方法[J]. 山东农业科学, 2025,57(11):149-158.
[7] 朱军, 陈钢, 蒋丽花, 等. 轻量化改进YOLOv8n的荔枝病虫害检测方法[J]. 农业工程学报, 2025, 41(13): 207-215.
[8] LIN T L, CHANG H Y, CHEN K H.The pest and disease identification in the growth of sweet peppers using faster R-CNN and mask R-CNN[J]. Journal of internet technology, 2020, 21(2): 605-614.
[9] BARI B S, ISLAM M N, RASHID M,et al.A real-time approach of diagnosing rice leaf disease using deep learning-based faster R-CNN framework[J]. Peerj computer science, 2021, 7: e432.
[10] DENG F M, MAO W, ZENG Z Q,et al.Multiple diseases and pests detection based on federated learning and improved faster R-CNN[J]. IEEE Transactions on instrumentation and measurement, 2022, 71: 1-11.
[11] GONG X L, ZHANG S J.A high-precision detection method of apple leaf diseases using improved faster R-CNN[J]. Agriculture, 2023, 13(2): 240.
[12] GUO W J, FENG S, FENG Q,et al.Cotton leaf disease detection method based on improved SSD[J]. International journal of agricultural and biological engineering, 2024, 17(2): 211-220.
[13] MAICAN E, IOSIF A, MAICAN S.Precision corn pest detection: Two-step transfer learning for beetles (coleoptera) with MobileNet-SSD[J]. Agriculture, 2023, 13(12): 2287.
[14] WANG J,YU L Y,YANG J,et al.DBA_SSD:A novel end-to-end object detection algorithm applied to plant disease detection[J]. Information, 2021, 12(11): 474.
[15] 魏志慧, 张聪, 成泞伸, 等. 一种水稻害虫的小目标检测方法研究[J]. 江苏农业科学, 2024, 52(9): 232-241.
[16] YUE X, QI K, NA X Y,et al.Deep learning for recognition and detection of plant diseases and pests[J]. Neural computing and applications, 2025, 37(17): 11265-11310.
[17] TIAN L L, ZHANG H X, LIU B,et al.VMF-SSD: A novel V-space based multi-scale feature fusion SSD for Apple leaf disease detection[J/OL]. IEEE/ACM Transactions on computational biology and bioinformatics, 2023, 20(3): 2016-2028.
[18] YIN Z B, LIU F Y, GENG H,et al.A high-precision jujube disease spot detection based on SSD during the sorting process[J]. Plos one, 2024, 19(1): e0296314.
[19] XUE Z Y, XU R J, BAI D,et al.YOLO-tea: A tea disease detection model improved by YOLOv5[J]. Forests, 2023, 14(2): 415.
[20] JIA L Q, WANG T, CHEN Y,et al.MobileNet-CA-YOLO: An improved YOLOv7 based on the MobileNetV3 and attention mechanism for rice pests and diseases detection[J]. Agriculture, 2023, 13(7): 1285.
[21] YU G W, MA B X, ZHANG R Y,et al.CPD-YOLO: A cross-platform detection method for cotton pests and diseases using UAV and smartphone imaging[J]. Industrial crops and products, 2025, 234: 121515.
[22] TIAN Y N, WANG S H, LI E,et al.MD-YOLO: Multi-scale Dense YOLO for small target pest detection[J]. Computers and electronics in agriculture, 2023, 213: 108233.
[23] MA Q, ZHAO W L, LIU Y L,et al.DLW-YOLO: An efficient algorithm for small target detection[J]. Journal of electronic imaging, 2025, 34(1): 13039.
[24] HUA J,HE R M, ZENG Y H,et al.HDMS-YOLO: A multi-scale weed detection model for complex farmland environments[J]. Frontiers in plant science, 2025, 16:1-16.
[25] HUI Y M, WANG J, LI B.WSA-YOLO: Weak-supervised and adaptive object detection in the low-light environment for YOLOV7[J]. IEEE Transactions on instrumentation and measurement, 2024, 73: 1-12.
[26] TIAN Y J, YE Q X, DOERMANN D. YOLOv12: Attention-centric real-time object detectors[EB/OL]. http://arxiv.org/abs/2502.12524, 2025-11-22.
[27] 李帅, 薄敬东, 龚瑞昆, 等. 基于多尺度特征增强的轻量化黄瓜病害识别模型[J]. 江苏农业科学,2024,52(20): 267-276.
[28] 朱悦云, 刘通, 何振荣. 基于多尺度特征增强的农作物病虫害识别[J]. 江苏农业科学, 2025, 53(9): 192-198.
[29] YANG H, LIN D, ZHANG G X,et al.Research on detection of rice pests and diseases based on improved yolov5 algorithm[J]. Applied sciences, 2023, 13(18): 10188.
[30] LI J, QIAO Y, LIU S,et al.An improved YOLOv5-based vegetable disease detection method[J]. Computers and electronics in agriculture, 2022, 202: 107345.
[31] SOEB M J A, JUBAYER M F, TARIN T A,et al. Tea leaf disease detection and identification based on YOLOv7 (YOLO-T)[J]. Scientific reports, 2023, 13(1): 6078.
[32] LUO D H, XUE Y J, DENG X R,et al.Citrus diseases and pests detection model based on self-attention YOLOV8[J]. IEEE Access, 2023, 11: 139872-139881.
[33] SUN H, NICHOLAUS I T, FU R,et al.YOLO-FMDI: A lightweight YOLOv8 focusing on a multi-scale feature diffusion interaction neck for tomato pest and disease detection[J]. Electronics, 2024, 13(15): 2974.
[34] HE R H, ZHANG F K, ZHU J K,et al.YOLOv9-LSBN: An improved YOLOv9 model for cotton pest and disease[J]. IEEE Access, 2025,13:136536-136553.
[35] ALZAHRANI A S, IQBAL A, ZAFAR W,et al.Swin-YOLO-SAM: A hybrid Transformer-based framework integrating Swin Transformer, YOLOv12, and SAM-2.1 for automated identification and segmentation of date palm leaf diseases[J]. Frontiers in plant science, 2025, 16: 1666374.
[36] 谢家兴, 廖飞, 王卫星, 等. 基于改进Faster R-CNN的荔枝病虫害检测[J]. 华中农业大学学报, 2025, 44(1): 62-73.
[37] 宋中山, 汪进, 郑禄, 等. 基于二值化的Faster R-CNN柑橘病虫害识别研究[J]. 中国农机化学报, 2022, 43(6): 150-158.
[38] 胡凯, 骆润玫, 刘泽乾, 等. 基于改进SSD的广佛手病虫害检测方法[J]. 南京农业大学学报, 2023, 46(4): 813-821.

融合多尺度特征与注意力机制的改进YOLOv12农作物虫害检测方法

Improved YOLOv12 with multi-scale feature fusion and attention mechanism for crop pest detection

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	霍静琦, 崔婷婷, 薛志璐. 基于改进YOLOv8模型的黄花菜花蕾识别研究[J]. 湖北农业科学, 2025, 64(7): 186-191.
[2]	班若楠, 董峦. 基于深度学习的盐碱地遥感影像反演模型构建[J]. 湖北农业科学, 2025, 64(6): 185-189.
[3]	何恒, 周平. 基于深度学习的三七病害识别方法适应性研究[J]. 湖北农业科学, 2024, 63(8): 54-60.
[4]	周橄尉, 陈嘉越, 吴佳伟, 赵雅琪, 赵奕凯, 张小英. 适用于嵌入式设备的轻量级深度学习网络RepYOLO[J]. 湖北农业科学, 2024, 63(8): 153-157.
[5]	白春晖, 陈健, 郜鲁涛. 基于改进BiSeNet的葡萄黑麻疹病害程度分级预测[J]. 湖北农业科学, 2024, 63(5): 187-193.
[6]	朱志贤, 邱盼, 张成, 董朝霞, 张凤, 胡兴明, 于翠. 基于YOLOv3深度学习算法的桑椹菌核病严重度检测方法研究与应用[J]. 湖北农业科学, 2024, 63(12): 191-198.
[7]	马海荣, 沈祥成, 罗治情, 陈娉婷, 郑明雪, 官波. 基于Psi-Net深度学习网络的高空间分辨率遥感影像地块尺度的耕地提取[J]. 湖北农业科学, 2024, 63(11): 197-202.
[8]	石珊, 胡兵, 杨丛瑞. 基于深度学习的建筑物识别及占用耕地建房自动化监测——以湖南省长沙市X村为例[J]. 湖北农业科学, 2024, 63(1): 195-198.
[9]	袁鹏, 王珂, 肖坚. 基于RMAU-Net网络模型的高分影像耕地提取[J]. 湖北农业科学, 2023, 62(8): 182-188.
[10]	李正, 李宝喜, 李志豪, 战艺芳, 王利华, 龚琦. 基于深度学习的农作物病虫害识别研究进展[J]. 湖北农业科学, 2023, 62(11): 165-169.
[11]	邱儒琼, 彭少坤, 李孟璠. 基于深度学习和遥感数据的水稻种植面积提取[J]. 湖北农业科学, 2023, 62(11): 176-182.
[12]	雷朦, 余顺园. 基于卷积神经网络的魔芋病害识别[J]. 湖北农业科学, 2022, 61(23): 197-201.
[13]	孙吉红, 钱晔, 周正, 张剑波. 深度学习环境下香料种植推荐模型的设计研究[J]. 湖北农业科学, 2022, 61(22): 174-177.
[14]	王培福, 孙一丹, 鹿子涵, 王伟, 陈晓峰. 基于深度学习的板栗分级方法研究[J]. 湖北农业科学, 2022, 61(21): 168-175.
[15]	陈行, 夏丽华, 颜军, 蒋晓旭, 黄腾杰, 邓剑文. 基于改进U-Net网络模型的高分影像养殖池塘提取[J]. 湖北农业科学, 2022, 61(20): 166-171.