Xu Sheng-He, Wang Sai
School of Mathematics and Statistics, Fuyang Normal University, Fuyang, China.
State Key Laboratory of Marine Resource Utilization in South China Sea, Hainan University, Haikou, Hainan, China.
Front Plant Sci. 2025 Aug 26;16:1608515. doi: 10.3389/fpls.2025.1608515. eCollection 2025.
The rapid development of intelligent technologies has transformed various industries, and agriculture benefits greatly from precision farming innovations. One of the remarkable achievements in agriculture is enhancing pest and disease identification for better crop health control and higher yields. This paper presents novel models of a multimodal data fusion technique to meet the growing need for accurate and timely wheat pest and disease identification. It combines image processing, sensor - derived environmental data, and machine learning for reliable wheat pest and disease diagnosis. First, deep - learning algorithms in image analysis detect early - stage pests and diseases on wheat leaves. Second, environmental data such as temperature and humidity improve diagnosis. Third, the data fusion process integrates image data for further analysis. Finally, several criteria compare the proposed model with previous methods. Experimental results show the proposed techniques achieve a detection accuracy of 96.5%, precision of 94.8%, recall of 97.2%, F1 score of 95.9%, MCC of 0.91, and AUC - ROC of 98.4%. The training time is 15.3 hours, and the inference time is 180 ms. Compared with CNN - based and SVM - based techniques, the proposed model's improvement is analyzed. It can be adapted for real - time use and applied to more crops and diseases.
智能技术的快速发展改变了各个行业，农业从精准农业创新中受益匪浅。农业的显著成就之一是加强病虫害识别，以更好地控制作物健康并提高产量。本文提出了一种多模态数据融合技术的新型模型，以满足对准确及时的小麦病虫害识别日益增长的需求。它结合了图像处理、传感器获取的环境数据和机器学习来进行可靠的小麦病虫害诊断。首先，图像分析中的深度学习算法检测小麦叶片上的早期病虫害。其次，温度和湿度等环境数据改善诊断。第三，数据融合过程整合图像数据以进行进一步分析。最后，通过几个标准将所提出的模型与以前的方法进行比较。实验结果表明，所提出的技术实现了96.5%的检测准确率、94.8%的精确率、97.2%的召回率、95.9%的F1分数、0.91的马修斯相关系数和98.4%的AUC-ROC。训练时间为15.3小时，推理时间为180毫秒。分析了所提出的模型与基于卷积神经网络（CNN）和支持向量机（SVM）的技术相比的改进情况。它可以适用于实时使用，并应用于更多作物和病害。
