摘要
人工智能在农业上的应用是目前的研究热点,在作物的高通量表型组学研究方面具有很好的应用前景。为了对种质资源和育种中间材料的表型进行精准化、智能化、高通量的采集,本研究将最新的目标检测算法和传统的图像处理方法相结合,将基于YOLOv5和图像处理的智能数据采集技术应用于菜用大豆荚型表型的识别。结果发现,该技术能够自动化、批量化地提取一张图片内单粒荚、双粒荚、三粒荚和四粒荚的个数,并获取这些豆荚的长宽数值。通过与实际的豆荚粒数进行对比发现,本研究方法的最大平均精度达98.96%以上,高于传统的深度学习分类网络;与实际的长宽数据进行对比,长宽决定系数均在95.23%以上。本研究所采取的基于深度学习的智能数据采集技术具有识别速度快、精准度高的优势,能大幅降低人工测量的时间成本和人力成本,提高品种选育的工作效率,为菜用大豆荚型的表型性状的高通量、智能化和精准化获取提供了一种新技术。
The application of artificial intelligence in agriculture is a frontier research hotspot curreutly and has good prospects for application in high-throughput phenomics research in crops.To perform accurate,intelligent,and high-throughput phenotyping of germplasm resources,the technique combine the state-of-art target detection algorithms and the corresponding image processing methods.Specifically,the technique use YOLOv5 to identify pod phenotypes of vegetable soybeans.The research shows that this technique can determine the number of single-,double-,triple-and quadruple-seed pods within a single image and fetch the corresponding length and width values automatically.The average accuracy of this technique achieves 98.96%.Meanwhile,compared with the actual length values,the determination coefficients are both above 95.23%.In general,the intelligent data collection technique based on deep learning has faster speed and higher accuracy in recognition.It can significantly reduce the time and labor cost and improve the work efficiency of plant breeding.It can also provide a new research tool for high-throughput,intelligent,and accurate morphology traits acquisition for vegetable soybean pods.
作者
翔云
陈其军
宋栩杰
蔡昌
卜远鹏
刘娜
XIANG Yun;CHEN Qijun;SONG Xujie;CAI Chang;BU Yuanpeng;LIU Na(College of Information Engineering,Zhejiang University of Technology,Hangzhou,Zhejiang 310000;State Key Laboratory for Managing Biotic and Chemical Threats to the Quality and Safety of Agro-products,Institute of Vegetables,Zhejiang Academy of Agricultural Sciences,Hangzhou,Zhejiang 310021)
出处
《核农学报》
CAS
CSCD
北大核心
2022年第12期2391-2399,共9页
Journal of Nuclear Agricultural Sciences
基金
浙江省重点研发计划项目(2021C02052)。
关键词
菜用大豆
深度学习
目标检测
图像处理
荚型
vegetable soybean
deep learning
object detection
image processing
pods morphology traits
