| 引用本文: | 李俊,朱尚尚,童毅,任前程,黄婕,胡宇,李佳佳.农业试验数据智能采集系统的设计与应用[J].中国农业信息,2022,34(5):40-53 |
| |
|
| 摘要: |
| 【目的】 为了快速准确获取农业科研试验中具有体量大、种类繁多且复杂、可变性大、对真实性有严格要求等特点的田间植株表型数据。【方法】 文章基于物联网与大数据技术构建了一个农业试验中大数据采集应用系统，用于辅助人工进行试验数据采集与应用。数据采集方法有基于NodeMcu开发板结合传感器获取环境数据、通过树莓派连接摄像头对试验区域进行图像采集并利用物联网设备上传至数据库、网络爬虫、通过终端设备记录作物单株形态等。获得数据后对不同来源数据进行清洗与处理，将原始数据与清洗处理后的数据分别存储至不同数据区域中并固化，通过分布式文件系统HDFS（Hadoop Distributed File System）读写操作，最后利用数据处理模块对数据进行监控与处理，将结果以图像、表格和视频等形式提交到前端交互网站。【结果】 基于构建的大数据采集系统获得了5 450幅大豆叶片图像，然后利用yolov5的深度学习模型训练，最终实现了大豆叶形分类识别；利用株高测量设备获取了1 306株大豆株高数据，结果较为可靠。【结论】 研究表明，该系统设计方案具有可行性高、用途广泛、构建成本低和可拓展性强等特点，将多种技术运用于农业试验的数据获取中，规范化试验流程与数据保存，提高数据获取的广度和数据利用的深度，为更深层次的农业科学研究奠定了基础。 |
| 关键词: 农业试验 田间植株表型 大数据采集 Hadoop |
| DOI：10.12105/j.issn.1672-0423.20220505 |
| 分类号: |
| 基金项目:国家重点研发计划（2021YFD1201603-4）；国家级大学生创新创业计划项目“农业试验中大数据采集与处理模型的构建与应用”（202110364053）；安徽省大学生创新创业计划项目“农业试验中大数据采集与处理模型的构建与应用”（S202010364128） |
|
| Design and application of intelligent data acquisition system for agricultural experiment |
|
Li Jun, Zhu Shangshang, Tong Yi, Ren Qiancheng, Huang Jie, Hu Yu, Li Jiajia
|
|
School of Agronomy，Anhui Agricultural University，Hefei 230036
|
| Abstract: |
| [Purpose] In order to quickly and accurately obtain field plant phenotypic data with large volume，variety and complexity，variability，and strict requirements for authenticity in agricultural research experiments.[Method] Based on the Internet of things and big data technology，this paper constructs a big data acquisition application system in agricultural experiment，which is used to assist manual test data acquisition and application. The Arduino development board was used to obtain the environmental data in agricultural experiments. The image of the test area was collected by the raspberry pie connected camera，and uploaded to the database based on the internet of things equipment. Finally，the plant morphology of crops was scanned by the terminal equipment，and then the data from different sources were cleaned and processed in different ways. The original data and the data after cleaning were stored in different data areas and solidified，and further operations were carried out through the distributed file system HDFS （Hadoop distributed file system）. The data processing module monitors and processes the data and submits the results to the front-end interactive website in the form of images，tables and videos.[Result] In this paper，5 450 soybean leaf images obtained by the system were used to train the deep learning model based on yolov5，and finally the classification and recognition of soybean leaf shape were realized. The plant height measurement equipment was used to measure the plant height of 1 306 soybean plants，and decent results were obtained.[Conclusion] The research shows that the design scheme of the system has the characteristics of high feasibility，wide application，low construction cost and strong expansibility.It applies various technologies to the data acquisition of agricultural experiments，standardizes the experimental process and data storage，improves the breadth of data acquisition and the depth of data utilization，and lays a foundation for deeper agricultural scientific research. |
| Key words: agricultural experiment field plant phenotype big data collection Hadoop |
