摘要
杂草是影响作物产量的主要生物因素之一,可阻碍作物的生长,进而影响作物的产量和品质,传统的人工除草、机械除草、化学药剂除草方式已不能满足精准除草的需求,而近年来基于卷积神经网络(CNN)的深度学习技术发展迅速,成为图像识别的重要工具,并且在杂草检测、害虫识别、植物或果实计数及其成熟度分级等农业领域取得了显著成果。为了使田间杂草管理更高效,促进农业生产智能化,基于卷积神经网络研究情况,从目标检测、图像分割、图像分类、基于无人机图像应用等方面概述了卷积神经网络在田间杂草管理中应用的研究进展,并从数据采集、杂草检测的精度、模型的泛化能力3个方面进行了展望。总之,基于CNN的杂草管理应用研究已经取得了一定的成果,但仍存在许多挑战和问题需要解决。未来研究应关注提高数据数量和质量、提高杂草识别的准确性和可靠性、提高深度学习模型的泛化能力和鲁棒性,并引导无人机自主进行制图,实现无人机与地面设备的协同作业;同时,加强CNN等深度学习技术在实际生产中多方面的应用研究,为农业生产提供更高效、智能的解决方案。
Weed is one of the main biotic factor that affects crop yield,which can hinder crop growth and subsequently affect yield and quality of crop.Traditional methods of manual weeding,mechanical weeding and chemical weeding can no longer meet the needs of precision weeding.In recent years,deep learning technology based on convolutional neural networks(CNN)has developed rapidly and has become an important tool for image recognition,whose significant achievements have been made in agricultural fields,such as weed detection,pest identification,plant or fruit counting and fruit maturity grading.In order to make weed management of field more efficient and promote the intelligence of agricultural production,based on the research status of CNN,the research progress of CNN in field weed management from the aspects of object detection,image segmentation,image classification,image applications on basis of unmanned aerial vehicle(UAV)were summarized,and prospects were made from three aspects of data collection,accuracy of weed detection and generalization ability of the mode.In summary,research on weed management applications based on CNN has achieved certain results,but there are still many challenges and problems that need to be addressed.Future research should focus on improving the quantity and quality of data,improving the accuracy and reliability of weed identification,enhancing the generalization ability and robustness of deep learning modes,and guiding UAV to autonomously perform mapping to achieve collaborative operations between UAV and ground equipment.At the same time,the application research of deep learning technology of CNN in various aspects of actual production should be strengthened to provide more efficient and intelligent solutions for agricultural production.
作者
张金梦
张倩
王明
谭雅蓉
陶震宇
于金莹
ZHANG Jinmeng;ZHANG Qian;WANG Ming;TAN Yarong;TAO Zhenyu;YU Jinying(Institute of Data Science and Agricultural Economics,Beijing Academy of Agriculture and Forestry,Beijing 100097,China)
出处
《蔬菜》
2024年第7期28-36,共9页
Vegetables
基金
北京市农林科学院青年基金(QNJJ202213)
北京市农林科学院改革与发展专项(GGFZSJS2024)。
关键词
卷积神经网络
杂草
管理
深度学习
检测
智能化
无人机
convolutional neural network
weed
management
deep learning
detection
intelligence
unmanned aerial vehicle(UAV)
