基于无人机数码影像的玉米育种材料株高和LAI监测  

Monitoring plant height and leaf area index of maize breeding material based on UAV digital images

作　　者：牛庆林 冯海宽[1,2,3] 杨贵军[1,2,3] 李长春 杨浩[1,2,3] 徐波 赵衍鑫[5] 

机构地区：[1]农业部农业遥感机理与定量遥感重点实验室,北京农业信息技术研究中心,北京100097 [2]国家农业信息化工程技术研究中心,北京100097 [3]北京市农业物联网工程技术研究中心,北京100097 [4]河南理工大学测绘与国土信息工程学院,焦作454000 [5]北京市农林科学院玉米研究中心,北京100097

出　　处：《农业工程学报》2018年第5期73-82,共10页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家重点研发计划(2016YFD0700303);国家自然科学基金(41601346,41601369);UAV机载高光谱遥感关键技术研究及在智慧农业中的应用(2016A002)

摘　　要：快速、无损和高通量地获取田间株高(height,H)和叶面积指数(leaf area index,LAI)表型信息,对玉米育种材料的长势监测及产量预测具有重要的意义。基于无人机(unmanned aerial vehicle,UAV)遥感平台搭载高清数码相机构建低成本的遥感数据获取系统,于2017年5—9月在北京市昌平区小汤山镇国家精准农业研究示范基地的玉米育种材料试验田,获取试验田苗期、拔节期、喇叭口期和抽雄吐丝期的高清数码影像和地面实测的H、LAI和地面控制点(ground control point,GCP)的三维空间坐标。首先,基于高清数码影像结合GCP生成试验田的数字表面模型(digital surface model,DSM)和高清数码正射影像(digital orthophoto map,DOM);然后,基于DSM和DOM分别提取玉米育种材料的H和数码影像变量,其中将DOM的红、绿和蓝通道的DN(digital number)值分别定义为R、G和B,进行归一化后得到数码影像变量,分别定义为r、g和b;最后,基于实测H对DSM提取的H进行了精度验证,并用逐步回归分析方法进行了LAI的估测。结果表明,实测H和DSM提取的H高度拟合(R^2、RMSE和n RMSE分别为0.93,28.69 cm和17.90%);仅用数码影像变量估测LAI,得到最优的估测变量为r和r/b,其估算模型和验证模型的R^2、RMSE和n RMSE分别为0.63,0.40,26.47%和0.68,0.38,25.51%;将H与数码影像变量进行融合估测LAI,得到最优的估测变量为H、g和g/b,其估算模型和验证模型的R^2、RMSE和n RMSE分别为0.69,0.37,24.34%和0.73,0.35,23.49%。研究表明,基于无人机高清数码影像结合GCP生成DSM,提取玉米育种材料的H,精度较高;将H与数码影像变量进行融合估测LAI,与仅用数码影像变量相比,估测模型和验证模型的精度明显提高。该研究可为玉米育种材料的田间表型信息监测提供参考。Acquiring high-throughput phenotypic information of crop height and leaf area index(LAI) in the fields rapidly and non-destructively is of great significance for monitoring growth of maize breeding material and predicting maize yield. Currently, phenotypic information of maize breeding material in the fields is obtained by traditional manual investigation, which is an inefficient, time-consuming work, as there are plenty of breeding material plots and there exists a certain degree of human subjectivity. Ultra-low altitude remote sensing data acquisition system based on unmanned aerial vehicle(UAV) platform with different remote sensing micro-sensors can acquire high-throughput crop phenotypic information fastly and non-destructively, overcoming the shortcomings of traditional field phenotypic information acquisition techniques, so it is becoming a research focus in crop phenotypic information technology. In this study, a low-cost UAV remote sensing data acquisition system equipped with a high-resolution digital camera was employed. Field phenotypic data of maize breeding material were acquired at the National Precision Agriculture Research and Demonstration Base in Xiaotangshan Town, Changping District, Beijing City from May to September in 2017. Three-dimensional coordinates of 16 ground control points(GCPs) evenly arranged on the ground were measured by a high-precision differential GPS(global positioning system). The high-resolution digital images of the digital camera were obtained at seedling, jointing, trumpet and anthesis-silking stages of maize. The average heights and LAI of maize in 72 randomly selected breeding plots were acquired almost synchronously with the flight campaigns. High-precision digital surface model(DSM) was produced based on high-resolution digital images of UAV and ground GCPs. Canopy heights of maize breeding material at each growth stage were obtained by calculating the differences of DSM between different growth stages. The maize heights derived from DSM and GCPs were verified in te

关 键 词：无人机 农作物 提取 数码影像 玉米育种材料 株高 叶面积指数 逐步回归 

分 类 号：S25[农业科学—农业机械化工程]