基于熵权-模糊综合评价法的无人机多光谱春玉米长势监测模型研究  

Growth Monitoring of Spring Maize Using UAV Multispectral Imaging Based on Entropy Weight-Fuzzy Comprehensive Evaluation Method

作　　者：赵经华[1,2] 马世骄 房城泰 ZHAO Jinghuaa;MA Shijiao;FANG Chengtai(College of Water Conservancy and Civil Engineering,Xinjiang Agricultural University,Urumqi 830052,China;Xinjiang Key Laboratory of Hydraulic Engineening Safety and Water Disaster Prevention,Urumqi 830052,China;Corps Soil and Water Conservation and Water Resources Development Center,Urumqi 830002,China)

机构地区：[1]新疆农业大学水利与土木工程学院,乌鲁木齐830052 [2]新疆水利工程安全与水灾害防治重点实验室,乌鲁木齐830052 [3]兵团水土保持与水利发展中心,乌鲁木齐830002

出　　处：《农业机械学报》2024年第8期214-224,共11页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(52169013);新疆维吾尔自治区十四五重大专项(2020A01003-4);自治区研究生科研创新项目(XJ2024G126)。

摘　　要：为实现春玉米长势的快速监测,实时掌握田间作物的生长状况,本文以新疆维吾尔自治区克拉玛依地区种植的春玉米作为研究对象,利用无人机多光谱影像对春玉米进行长势监测。基于地面采集的春玉米叶片叶绿素含量、叶面积指数、地上部生物量和株高等数据,结合熵权法(EWM)和模糊综合评价法(FCE)建立综合长势指标CGMI_(EWM)和CGMI_(FCE)。通过无人机遥感影像数据构建光谱指数,并利用皮尔逊相关性分析法和方差膨胀因子确定模型最佳输入变量。采用偏最小二乘法(PLS)、随机森林回归(RF)及粒子群算法(PSO)优化RF模型建立春玉米长势反演模型,结合模型精度评价指标,最终确定春玉米空间影像长势分布图。结果表明,以CGMI_(EWM)和CGMI_(FCE)构建综合长势指标的相关性均高于单一长势指标的相关性;利用CGMI_(FCE)长势指标结合PSO-RF模型反演春玉米长势的效果最优,其决定系数(R^(2))为0.823,均方根误差(RMSE)为0.084%,相对分析误差(RPD)为2.345;研究区春玉米长势集中在生长正常(ZZ)等级,说明全区春玉米长势较为稳定。研究结果可为春玉米的田间管理提供科学依据。To achieve rapid monitoring of spring maize growth and gain real-time understanding of field crop conditions,focusing on spring maize planted in the Karamay region of Xinjiang,utilizing UAV multispectral imagery for growth monitoring of the spring maize,based on ground-collected data on spring maize leaf chlorophyll content,leaf area index,aboveground biomass,and plant height,comprehensive growth indicators CGMI_(EWM)and CGMI_(FCE)were established by combining the entropy weight method(EWM)and fuzzy comprehensive evaluation(FCE).Spectral indices were constructed by using UAV remote sensing imagery data,and the optimal input variables for the model were determined by using Pearson correlation analysis and the variance inflation factor.Partial least squares(PLS),random forest regression(RF),and particle swarm optimization(PSO)were used to optimize the RF model and establish a spring maize growth inversion model.By combining model accuracy evaluation metrics,the spatial distribution map of spring maize growth was ultimately determined.The results showed that the comprehensive growth indicators constructed using CGMI_(EWM)and CGMI_(FCE)had higher correlations than single growth indicators.The growth indicators derived from CGMI_(FCE),combined with the PSO-RF model,resulted in the best performance for inversion of spring maize growth.The coefficient of determination(R^(2))was 0.823,the root mean square error(RMSE)was 0.084%,and the relative percent deviation(RPD)was 2.345.The growth of spring maize in the study area was mostly concentrated in the normal growth(ZZ)category,indicating relatively stable growth across the region.The research results can provide a scientific basis for the field management of spring maize and offer a data foundation for the development of precision agriculture.

关 键 词：春玉米 熵权法 模糊综合评价法 综合长势指标 多光谱 无人机监测 

分 类 号：S127[农业科学—农业基础科学]