"Research on Drought Stress Detection in the Seedling Stage of Yunnan Large-Leaf Tea Plants Based on Biomimetic Vision and Chlorophyll Fluorescence Imaging Technology"

"To address the problem of drought stress detection in the seedling stage of Yunnan large-leaf tea plants, this study proposes an improved network, MC-YOLOv13-L, based on YOLOv13. With the compound eye's parallel sampling mechanism at its core, Compound-Eye Apposition Concatenation optimization is applied in both the training and inference stages. Simulating the environmental information acquisition and integration mechanism of primates\u2019 \"multi-scale parallelism\u2014global modulation\u2014long-range integration,\" multi-scale linear attention is used to optimize the network. Simulating the retinal wide-field lateral inhibition and cortical selective convergence mechanisms, CMUNeXt is used to optimize the network's backbone. To further improve the localization accuracy of drought stress detection and accelerate model convergence, a dynamic attention process simulating peripheral search, saccadic focus, and central fovea refinement in primates is used. Inner-IoU is applied for targeted improvement of the loss function. The experimental results show that the Box Loss, Cls Loss and DFL Loss of MC-YOLOv13-L on the training set are decreased by 5.08 %, 3.13 % and 4.85 % respec-tively compared with YOLOv13, and decreased by 2.82 %, 7.32 % and 3.51 % respectively on the verification set. On the whole, the improved MC-YOLOv13-L improves the accuracy, recall rate and mAP by 4.64 %, 6.93 % and 4.2 % respectively on the basis of only sacrificing 0.63 FPS. The network can not only quickly and accurately identify the drought stress response of tea plants, but also provide a reliable technical basis for the intelligentization of tea production, and also provide a useful reference for the practical application and transformation of bio-heuristic computing in complex ecosystems."

针对云南大叶茶苗期干旱胁迫检测难题，本研究提出一种基于YOLOv13改进的网络MC-YOLOv13-L。该网络以复眼并行采样机制为核心，在训练与推理阶段均采用复眼并列串联（Compound-Eye Apposition Concatenation）优化方案。模拟灵长类动物「多尺度并行—全局调制—远程整合」的环境信息获取与整合机制，采用多尺度线性注意力对网络进行优化；同时模拟视网膜广域侧抑制与皮层选择性汇聚机制，采用CMUNeXt对网络主干进行优化。为进一步提升干旱胁迫检测的定位精度并加速模型收敛，本研究引入模拟灵长类动物外周搜索、扫视聚焦与中央凹精细化过程的动态注意力机制，并针对损失函数进行针对性改进，采用内部交并比（Inner-IoU）。实验结果表明，相较于原始YOLOv13，MC-YOLOv13-L在训练集上的边界框损失（Box Loss）、分类损失（Cls Loss）与分布焦点损失（DFL Loss）分别下降5.08%、3.13%与4.85%，在验证集上则分别下降2.82%、7.32%与3.51%。整体而言，改进后的MC-YOLOv13-L仅以0.63 FPS的帧率损失为代价，将准确率、召回率与平均精度均值（mean Average Precision）分别提升4.64%、6.93%与4.2%。该网络不仅能够快速准确地识别茶树干旱胁迫响应，可为茶叶生产智能化提供可靠的技术支撑，同时也为生物启发计算在复杂生态系统中的实际应用与转化提供了有益参考。