叶面积指数（leaf area index，LAI）是表征作物生长信息的重要参数，利用无人机遥感平台获取农作物光谱信息定量反演LAI对精确监测作物生长情况具有重要意义。本文以玉米为研究对象，利用无人机（unmanned aerial vehicle，UAV）搭载MicaSense RedEdge-M多光谱成像仪获取玉米拔节期、抽雄期、成熟期等关键生育期内低空遥感影像，同步采集地面LAI，基于多光谱信息构建植被指数并研究其与LAI的定量关系，进一步建立玉米LAI估算模型，对比分析筛选最优植被指数与最适监测时期。实验发现在拔节期、抽雄期、成熟期玉米LAI与NDVI、OSAVI、EVI、NDRE均具有较好的相关性；在不同时期分别基于OSAVI、NDRE、NDRE建立了LAI监测模型，模型监测精度分别为0.549、0.753、0.733；验证模型精度分别为0.907、0.932、0.926，模型估算值与田间实测值间相对误差分别为8.57、8.37、9.24，均方根误差分别为0.104、0.087、0.091；基于不同生育时期LAI估算模型进行田块尺度的LAI空间分布制图，估算值与实测值的决定系数分别为0.883、0.931、0.867；相对误差分别为：9.17、8.86、9.32。结果表明基于MicaSense RedEdge-M多光谱成像仪能有效估算玉米关键生育时期LAI，可为定量实时估算田块尺度的玉米LAI提供理论依据。

Remote sensing technology can be used to estimate the leaf area index (LAI) value of crops rapidly and harmlessly. The purpose of this study is to research the accuracy, reliability, and adaptability of the LAI using unmanned aerial vehicle (UAV) multispectral remote sensing. During a summer maize-fertilizer cross test, the LAI and multispectral images captured by a six-rotor UAV with a MicaSense RedEdge-M camera (which has five high-resolution channels: blue, green, red, red edge, and near infrared) were collected at the jointing, tasseling, and maturity stages of the maize. The normalized differential vegetation index (NDVI), optimized soil adjusted vegetation index (OSAVI), enhanced vegetation index (EVI), and normalized differential red edge index (NDRE) were calculated at each stage. The correlation between these metrics and the LAI were analyzed and their values were established based on the multispectral images at different growth stages. Then, an LAI model for each growth stage was established. After the accuracy of these models was tested using independent data, a maize LAI estimation map was made by processing each pixel in the maize multispectral image using these models. The results indicate the following: 1) There is a high correlation between the LAI and the NDVI, OSAVI, EVI, and NDRE values at the jointing, tasseling, and maturity stages. 2) LAI estimation models were established based on OSAVI, NDRE, and NDRE for the jointing, tasseling, and maturity stages, respectively. They had decision coefficient values (R²) of 0.549, 0.753, and 0.733, respectively, and the R² of the verification models were 0.907, 0.932, and 0.926, respectively. The predicted and measured values at different growth stages had relative error values of 8.57, 8.37, and 9.24 and root-mean-squared error values of 0.104, 0.087, and 0.091, respectively. 3) The spatial distribution of the LAI at field scale was mapped by the LAI estimation models at each growth stage, yielding R² values of 0.883, 0.931, and 0.867 and relative error values of 9.17, 8.86, and 9.32, respectively. Therefore, the LAI map reflected the real-world spatial distribution pattern of the LAI in the maize fields well. The established agricultural UAV remote sensing monitoring system provides accuracy, reliability, and adaptability for precision agriculture applications as well as the corresponding retrieval models for studying the feasibility of estimating the LAI during different growth stages.