基于光谱指数的棉花黄萎病叶片氮素含量提取  

Extracting Nitrogen Contents in Leaves of Cotton under Verticillium Wilt Stress Based on Hyper Spectra

作　　者：陈兵[1] 邓福军[1] 林海[1] 韩焕勇[1] 王方永[1] 刘政[1] 

机构地区：[1]新疆农垦科学院棉花研究所/农业部西北内陆区棉花生物学与遗传育种重点实验室/国家棉花改良中心新疆生产建设兵团分中心,新疆石河子832003

出　　处：《新疆农业科学》2012年第12期2222-2228,共7页Xinjiang Agricultural Sciences

基　　金：国家自然科学基金(41161068);新疆农垦科学院科技引导计划(YYD201102)

摘　　要：【目的】研究棉花黄萎病叶片氮素含量与高光谱的关系,以期用简便、无损的遥感技术提取病害棉叶氮素含量,为大面积遥感监测棉花病害提供理论依据。【方法】通过小区和大田同步调查棉花黄萎病,在不同生育期测定病叶光谱及其氮素含量。将病叶光谱特征参数与氮素含量进行相关分析,建立病叶氮素含量估测模型并检验。【结果】随着病害严重度的增加,棉叶氮素含量逐渐减小。病叶氮素含量与光谱指数FD731、NDVI[670,890]、DVI[FD554,FD731]、PVI[FD554,FD731]、RDVI[702,758]、RDVI[FD554,FD731]、SAVI、OSAVI、PRI[570,531]、PRI[702,758]、REP、Lo、Depth672和Area672呈极显著正相关,与11550、R680、R702、SD737、DVI[4So,560]、NDVI[702,758]、DVI[702,758]、RVI[702,758]、SIPI、TCARI、CCII、PPR[550,450]、Lwidth和ND672均呈极显著负相关,与Dr未达显著相关。选取相关系数较大的光谱参数建立的病叶氮素含量估测模型均达到显著水平,整体上利用DVI[702,758]、PVI[FD554,FD731]和NDVI[702,758]进行氮素含量的估测精度最高,模型的预测的相对误差均小于2%。【结论】考虑到DVI[702,758]建立的模型更为实用,可作为病害棉叶氮素含量的最佳估测模型。[ Objective ] The purpose of this project was to extract leaves nitrogen contents （LNC） of cotton under Verticillium wilt stress based on simple and undamaged hyper spectra technology in order to provide theory reference for monitoring large area cotton disease with remote sensing by studying the relations of nitrogen contents in leaves of cotton under Verticillium wilt stress and hyper spectra. [ Method ] The spectrum reflectance and LNC of cotton infected by Verticillium wilt ware measured in cotton disease nursery and field in different growth phases, and severity level of Verticillium wilt was investigated. The correlation between LNC of diseased cotton and spectral indices were analyzed respectively. The estimation models of LNC of diseased cotton were established and tested. [ Result ] The result revealed that LNC of cotton decreased little by little with the worsening disease. LNC had best significant positive correlations with spectral indices of FD731,NDVI [670,890], DVI [FD554, FD731 ], PVI[FD554, FD731 ], RDVI [702,758], RDVI [ FD554, FD731 ], SAVI, OSAVI, PRI [570,531 ], PRI [702,758], REP, Lo, Depth 672 and Area 672 had best significant negative correlations with R550, R680, R702, SD737, DVI [ 450,560 ], NDVI [ 702, 758], DVI [702,758], RVI [702,758], SIPI, TCARI, CCII, PPR [550,450], Lwidth and ND672,and no best significant correlations with Dr. The models which were selected from spectral indices had all attached significant correlation. The tested results indicated that the model of DVI [ 702,758 ], PVI [ FD554, FD731 ] and NDVI [ 702,758 ] had best estimated precision, and the relative errors were all within 2%. [ Conclusion ] When considering that the model of DVI [ 702,758 ] is very simple and practical, it was commended as a best model to estimate LNC of the diseased cotton.

关 键 词：棉花 病害胁迫 高光谱 氮素含量 估测模型 

分 类 号：S562[农业科学—作物学] S188