多尺度异构特征协同的棉花叶片病害检测模型OA
Multi-Scale Heterogeneous Feature Synergistic Model for Cotton Leaf Disease Detection
[目的/意义]面对自然田间场景中背景干扰强、病斑多尺度分布,以及边缘设备算力受限等多重挑战,现有的农业计算机视觉应用亟需高效的解决方案.基于改进RT-DETR(Real-Time Detection Transformer)架构,提出了多尺度异构特征协同的棉花叶片病害检测模型(Multi-Scale Heterogeneous Synergistic Feature DETR,MHSF-DETR).[方法]其核心在于重构了特征提取与融合的内在机制.该模型构建了分层上下文选择感知主干网络,在浅层特征阶段嵌入微宏观空间上下文注意力,防止微小病斑细节在深层网络中被淹没;在深层特征阶段集成竞争性选择融合模块,提升语义定位准确性;在特征融合层面,引入可学习加权上下文融合模块,通过动态学习权重解决层级间的语义错位,并结合边界感知重构机制,以反射填充和部分卷积策略有效抑制了边缘伪影,从而在保证特征鲁棒性的同时削减计算开销.[结果和讨论]消融实验验证了各模块的结构协同效应.与基准模型相比,该模型参数量与计算量分别降低22.42%和13.29%,平均精度均值(mAP50)提高了3.2个百分点.与主流算法横向对比,该模型实现精度与效率的更优平衡;在外部数据集的泛化测试下,其平均精度均值分别可达 80.3%与86.9%,跨域检测精度优于同规模主流算法.此外,同等条件的边缘仿真测试中,该模型推理速度最优,单帧耗时约1454 ms,具备低功耗部署潜力.上述结果证实,该模型的分层特征提取与动态融合机制的协同作用效果显著,架构层面的优化可在低算力约束下实现精度与效率的统一.[结论]本研究提出的模型MHSF-DETR能够有效执行复杂场景下的低功耗检测任务,为构建高效的田间实时监测系统提供技术支撑.
[Objective]Detecting cotton leaf diseases in natural field environments is challenging due to diverse image interferences,variable leaf-spot̂ sizes,and t̂he demand for fast̂ comput̂at̂ion on mobile phones and ot̂her small-scale devices.Nevertheless,existing lightweight models cannot well balance detection accuracy and computational efficiency,especially for detecting small lesions and suppressing noise surrounding leaves.To address these challenges,MHSF-DETR(Mult̂i-Scale Het̂erogeneous Synergist̂ic Feat̂ure DE-TR),an improved det̂ect̂ion model based on t̂he RT-DETR framework,is proposed.It̂ aims t̂o achieve high-precision,low-power diag-nosis in complex agricultural scenarios.[Methods]The primary innovation of this study consisted of the complete reconfiguration of the feature extraction and fusion architectures.Firstly,a hierarchical context-selective perception network(HCSP-Net)was construct-ed as the backbone to replace conventional architectures.This backbone employed a differentiated processing strategy tailored to the depth of the feature maps:In the early parts of the process where the features were simple,it used something called micro-macro spa-tial context attention(M2-SCA).This module used a channel semantic filter then a dual stream spatial perception structure to actively capture high frequency textures of micro-lesions and preserve macro semantics so that fine details were not lost when downsampled.At the deep feature stage,a competitive selection fusion(CSF)module was added.Unlike the traditional static summing approach,CSF created a dynamic competition arbitration system that flexibly balances local importance versus overall coherence via soft compe-tition gates,making the semantics sharper and filtering away irrelevant background noise.Secondly,to tackle the spatial and semantic misalignment that was commonly seen in cross-level feature fusions,a learnable weighted context fusion(LWC-Fusion)module was created inside the neck network.This module used global amplitude dynamic weighting to learn autonomously the best blending ra-tios,so that deep semantic features were aligned precisely with shallow geometry.Moreover,to solve the problem of artifacts appear-ing at irregular leaf boundaries caused by traditional zero-padding convolutions,an edge-aware reconstruction mechanism(EARM)was proposed.By using edge-refined convolution(ER-Conv)and the edge-refined convolution C3 module(ER-ConvC3),which inte-grated reflection padding and partial convolution techniques,the model successfully curtailed invalid edge noise and diminished com-putational redundancy without sacrificing the geometric continuity of features.[Results and Discussions]Empirical benchmarks demon-strated that the proposed MHSF-DETR achieved a superior balance between detection performance and computational efficiency.Compared to the RT-DETR-R18 baseline,MHSF-DETR yielded a significant 3.2 percentage points increase in mean average preci-sion(mAP),while simultaneously reducing parameters by 22.42%and GFLOPs by 13.29%.When benchmarked against mainstream detectors,MHSF-DETR consistently outperformed models such as YOLOv5m,YOLOv10m,and RT-DETR-R50.Although YO-LOv8m maintained a marginal mAP50 lead in specific scenarios,its exorbitant computational overhead rendered it less practical for real-time deployment compared to MHSF-DETR.It successfully matched the lean efficiency of YOLOv10m but excelled in detection accuracy.Furthermore,extensive ablation studies confirmed that these performance gains stemmed from the structural synergy among the HCSP-Net backbone,the LWC-Fusion neck,and specialized reconstruction modules,rather than isolated component upgrades.These results validated the effectiveness of MHSF-DETR design in optimizing feature extraction and fusion,offering a highly effi-cient solution for resource-constrained object detection tasks.[Conclusions]MHSF-DETR addresses t̂he long-st̂anding accuracy-efficiency t̂rade-off in cot̂t̂on disease monit̂oring.By integrating hierarchical perception,adaptive fusion and edge refine-ment,t̂he model mit̂igat̂es scale disparit̂y and resource-const̂raint̂ challenges.It̂ provides a feasible light̂weight̂ t̂emplat̂e for real-t̂ime diagnosis on agricult̂ural edge devices t̂o support̂ deployment̂ in smart̂-farming syst̂ems.Future work will expand validation to other plant organs(bolls,stems)and conduct̂ rigorous embedded-hardware field t̂est̂s t̂o assess real-world robust̂ness.
沈学利;张越;金海波;张旭旭
辽宁工程技术大学软件学院,辽宁 葫芦岛 125105,中国辽宁工程技术大学软件学院,辽宁 葫芦岛 125105,中国辽宁工程技术大学软件学院,辽宁 葫芦岛 125105,中国辽宁工程技术大学软件学院,辽宁 葫芦岛 125105,中国
信息技术与安全科学
棉花病害检测RT-DETR轻量化模型注意力机制特征融合
cotton disease detectionRT-DETRlightweight modelattention mechanismfeature fusion
《智慧农业(中英文)》 2026 (4)
204-216,13
国家自然科学基金项目(62173171) National Natural Science Foundation of China(62173171)
评论