深度学习在蔬菜生长监测中的应用研究进展OA
Research Progress on the Application of Deep Learning in Vegetable Growth Monitoring
为全面了解深度学习技术在蔬菜生长监测中的应用进展,基于大量文献调研,重点围绕基于计算机视觉模型的监测技术和基于时序和多模态数据的监测技术两大研究方向展开梳理与分析.研究表明,深度学习技术,特别是卷积神经网络(Convolutional Neural Networks,CNN)、循环神经网络(Recurrent Neural Networks,RNN)、Transformer模型以及结合图像和传感器数据的多模态深度学习方法在蔬菜生长指标无损预测、病虫害识别、产量预测、生理指标检测等方面表现出色,显著提升了监测的自动化与精准化水平.然而,该技术仍面临数据质量波动大、标注成本高、模型迁移能力差、边缘设备计算资源有限及模型泛化能力不足等挑战.针对这些问题,提出了多源数据融合、弱监督学习、模型跨域部署、模型轻量化与边云协同、多任务学习等解决策略.未来,随着自动化数据采集、跨模态学习与智能调控技术的发展,深度学习将在蔬菜生长监测与智慧农业管理中发挥更重要的作用.
To comprehensively understand the application progress of deep learning technology in vegetable growth monitoring,this paper systematically reviews and analyzes two major research directions:monitoring techniques based on computer vision models and those based on temporal and multimodal data,based on extensive literature research.Studies have shown that deep learning technologies,particularly convolutional neural networks(CNN),recurrent neural networks(RNN),Transformers model,and multimodal deep learning methods combining image and sensor data,have demonstrated outstanding performance in non-destructive estimation of vegetable growth indicators,pest and disease identification,yield prediction and physiological index detection,significantly enhancing the automation and precision of monitoring.However,challenges such as significant fluctuations in data quality,high annotation costs,poor model transferability,limited computational resources on edge devices and insufficient model generalization capabilities remain.To address these issues,strategies such as multi-source data fusion,weakly supervised learning,cross-domain model deployment,model lightweighting and edge-cloud collaboration,as well as multi-task learning have been proposed.In the future,with advancements in automated data acquisition,cross-modal learning,and intelligent regulation technologies,deep learning is expected to play an even more critical role in vegetable growth monitoring and smart agricultural management.
胡月;缪祎晟;吴华瑞
北京市农林科学院信息技术研究中心,北京 100097||国家农业信息化工程技术研究中心,北京 100097||农业农村部数字乡村技术重点实验室,北京 100097||江苏大学计算机科学与通信工程学院,江苏镇江 212013北京市农林科学院信息技术研究中心,北京 100097||国家农业信息化工程技术研究中心,北京 100097||农业农村部数字乡村技术重点实验室,北京 100097北京市农林科学院信息技术研究中心,北京 100097||国家农业信息化工程技术研究中心,北京 100097||农业农村部数字乡村技术重点实验室,北京 100097
农业科技
深度学习蔬菜生长监测计算机视觉时间序列预测多模态融合卷积神经网络Transformer模型
deep learningvegetablegrowth monitoringcomputer visiontime series predictionmultimodal fusionconvolutional neural network(CNN)Transformer model
《蔬菜》 2026 (8)
1-12,12
国家重点研发计划项目(2024YFD2000805)国家现代农业产业技术体系项目(CARS-21-D07)北京市智慧农业创新团队项目(BAIC10-2025).
评论