中国农业气象

• 论文 • 上一篇    

基于HJ-1A/B CCD数据的玉米倒伏识别方法

王杰,刘实,兰玉彬,陈立文,郭永青,王颖   

  1. 1. 山东理工大学国际精准农业航空应用技术研究中心,淄博 255000;2. 山东理工大学交通与车辆工程学院,淄博 255000; 3. 吉林省气象局,长春 130062;4. 山东理工大学农业工程与食品科学学院,淄博 255000;5. 吉林大学地球探测科学与技术学院,长春 130026;6. 吉林省气象科学研究所,长春 130062
  • 出版日期:2020-02-20 发布日期:2020-03-20
  • 作者简介:王杰,E-mail:wang-jie@ sdut.edu.cn
  • 基金资助:
    山东省引进顶尖人才“一事一议”专项经费资助项目(2018.01?2021.12);中央引导地方科技发展专项资金“精准农业航空技术与装备研发”资助项目(2017.1?2019.12);淄博市科技发展计划资助项目(2018kj010073);吉林省科技发展计划项目(20140204052SF)

Method of Maize Lodging Recognition Based on HJ-1A/B CCD Data

WANG Jie, LIU Shi, LAN Yu-bin, CHEN Li-wen, GUO Yong-qing, WANG Ying   

  1. 1.International Research Center of Precision Agricultural Aviation Application Technology, Shandong University of Technology, Zibo 255000, China; 2. School of Transportation and Vehicle Engineering, Shandong University of Technology, Zibo 255000; 3. Meteorological Service of Jilin Province,Changchun 130062; 4. School of Agricultural Engineering and Food Science, Shandong University of Technology, Zibo 255000; 5. College of Geo-exploration Science and Technology, Jilin University, Changchun 130026; 6.Jilin Provincial Institute of Meteorological Science, Changchun 130062
  • Online:2020-02-20 Published:2020-03-20

摘要: 为快速获取大面积玉米倒伏灾情信息,以2012年台风“布拉万”过境导致大面积玉米倒伏的公主岭市为研究区,利用HJ-1A/BCCD数据,对受灾前后倒伏玉米和正常玉米之间的光谱差异进行分析,提取归一化植被指数(NDVI)、比值植被指数(RVI)、增强植被指数(EVI)、差值植被指数(DVI)及4波段光谱反射率主成分,结合地面调查构建基于二元Logistic回归的玉米倒伏识别模型,并进行精度评价和验证。结果表明:玉米倒伏后冠层光谱反射率在可见光/近红外波段均表现为增大,但植被指数减小;二元Logistic回归方法对玉米倒伏识别适用,所建模型中以4波段光谱反射率主成分构建的二元Logistic回归模型对玉米倒伏的识别效果最优,测试集上分类结果的准确率达到96.23%,NDVI和RVI模型次之,准确率为80%左右;将主成分模型应用于公主岭市倒伏玉米识别,结果与灾情实际情况基本一致。基于二元Logistic回归模型对玉米倒伏进行监测的思路和方法可为区域尺度玉米倒伏的多光谱遥感监测提供参考。

关键词: 遥感, 环境减灾卫星, 灾害, 玉米, 倒伏

Abstract: To quickly and effectively obtain crop lodging information, this study proposed a remote sensing method for monitoring maize lodging using HJ-1(Small Satellite Constellations for Environment and Disaster Monitoring and Forecasting) CCD data. This paper took one area in Gongzhuling, Jilin Province as an example, where large-scale maize lodging occurred in 2012, caused by Typhoon Bolaven. The spectral characteristics of lodging and normal maize were first analyzed and summarized before and after the Typhoon. The results showed that compared with the normal field, the canopy reflectance increased in chromatic and near-infrared bans, but the vegetable index decreased in the lodged field. Four vegetation indices and a principal component were calculated, which extracted from 4 bans spectral data set. Binary Logistic models of lodging and normal maize classification were constructed with these 5 varieties. The prediction accuracies of the classification models were measured by ground survey samples. The principal component model could get the optimal results of recognition, and the classification accuracy on the test set was 96.23%. The classification accuracies of NDVI model and RVI model followed, the classification accuracies were about 80%. Finally, the principal component model was applied to recognize maize lodging using the spectral image, and the results confirmed that the proposed modes can accurately predict the distribution of maize lodging. The proposed maize lodging recognition method based on binary Logistic, provided a theoretical basis for monitoring large-scale lodged maize filed using multispectral data.

Key words: Remote sensing, Small satellite constellations for environment and disaster monitoring and forecasting, Disaster, Maize, Lodging