摘要
基于无人机数码影像,探讨融合光谱信息与纹理特征构建的"图-谱"指标对冬小麦氮营养指数的估算能力,为冬小麦氮素营养精准探测提供一种可靠的技术手段。利用无人机数码影像及相应的生物量和植株氮含量数据,分析了图像指数、纹理特征与氮营养指数的相关性,然后将图像指数与纹理特征相乘或相除融合形成"图-谱"融合指标,分析"图-谱"融合指标与氮营养指数的相关性,整合灰色关联度和方差膨胀因子,筛选对氮营养指数敏感的"图-谱"融合指标,最后用偏最小二乘法分析图像指数、纹理特征及"图-谱"融合指标估测氮营养指数的能力。结果表明:"图-谱"融合指标较图像指数、纹理特征与氮营养指数的相关性有了较大的提高,利用"图-谱"融合指标构建的氮营养指数模型估算精度(R^2=0.644 3)高于图像指数及纹理特征构建的氮营养指数模型(R^2分别为0.593 8及0.584 5),而且"图-谱"融合指标构建的模型验证结果均方根误差最小,为0.114 0。基于光谱信息和纹理特征融合的"图-谱"指标可以有效提高冬小麦氮营养指数的反演精度,为冬小麦氮素营养诊断反演提供了一种有效的思路。
Based on digital images of unmanned aerial vehicle,the effects of "image-spectrum" fusion indexes based on spectral information and texture features on the estimation of nitrogen nutrition index in winter wheat were investigated,which provided a new method for accurate estimation of nitrogen nutrition status.Digital images of unmanned aerial vehicle and corresponding biomass and nitrogen content data were used.The correlation between image indexes,texture features and nitrogen nutrition index was firstly analyzed,and then the image indexes and texture features were multiplied or divided and fused to form "image-spectrum" fusion indexes,the correlation was also analyzed between "image-spectrum" fusion indexes and nitrogen nutrition index,and "image-spectrum" fusion indexes which sensitive to nitrogen nutrition index were selected based on the integration of grey relation analysis and variance inflation factor.Finally,the ability of image indexes,texture features and "image-spectrum" fusion indexes to estimate nitrogen nutrition index was estimated by partial least square analysis.The results showed that the correlation between "image-spectrum" fusion indexes and nitrogen nutrition index had been greatly improved than the correlation between image indexes,texture features and nitrogen nutrition index.The estimation accuracy of the nitrogen nutrition index model constructed by the "image-spectrum" fusion indexes (R^2 equal to 0.644 3) was higher than the models constructed by image indexes (R^2 equal to 0.593 8) and texture features (R^2 equal to 0.584 5),and the verification result of model which constructed by the"image-spectrum" fusion indexes had the smallest root mean square error of 0.114 0.The "image-spectrum" indexes basedon the fusion of spectral information and texture features could effectively improve the inversion accuracy of the winter wheatnitrogen nutrition index,and provided an effective idea for the inversion of winter wheat nitrogen nutrition diagnosis.
作者
杨福芹
冯海宽
肖天豪
李天驰
郭向前
YANG Fu-qin;FENG Hai-kuan;XIAO Tian-hao;LI Tian-chi;GUO Xiang-qian(College of Civil Engineering,Henan University of Engineering,Zhengzhou,Henan 451191,China;National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China;Henan Surveying and Mapping Engineering Institute,Zhengzhou,Henan 450003,China;Institute of Surveying,Mapping and Geoinformation of Henan Provincial Bureau of Geo-exploration and Mineral Development,Zhengzhou,Henan 450006,China)
出处
《农业现代化研究》
CSCD
北大核心
2020年第4期718-726,共9页
Research of Agricultural Modernization
基金
国家自然科学基金项目(41601346)
2020年度河南省科技攻关计划项目(202102310333)
河南省高等学校重点科研项目计划(19A420006)。
关键词
无人机
数码影像
纹理特征
氮营养指数
“图-谱”融合指标
unmanned aerial vehicle(UAV)
digital images
texture features
nitrogen nutrition index
"image-spectrum"fusion indexes
