本文選題：遙感　切入點(diǎn)：提取　出處：《農(nóng)業(yè)工程學(xué)報》2017年07期

【摘要】：準(zhǔn)確掌握農(nóng)作物的空間種植分布情況,對于國家宏觀指導(dǎo)農(nóng)業(yè)生產(chǎn)、制定農(nóng)業(yè)政策有重要意義。針對黑龍江省玉米與大豆生育期接近、光譜特征相似,較難區(qū)分的問題,以多時相16 m空間分辨率高分一號(GF-1)衛(wèi)星寬覆蓋(wide field of view,WFV)影像為數(shù)據(jù)源,選擇歸一化植被指數(shù)(normalized difference vegetation index,NDVI)、增強(qiáng)植被指數(shù)(enhanced vegetation index,EVI)、寬動態(tài)植被指數(shù)(wide dynamic range vegetation index,WDRVI)、歸一化水指數(shù)(normalized difference water index,NDWI)4個特征,結(jié)合實(shí)地調(diào)查樣本點(diǎn),采用隨機(jī)森林分類算法,提取黑龍江省黑河市嫩江縣玉米與大豆種植面積。研究表明,區(qū)分玉米與大豆的最佳時段為9月下旬至10月上旬,即大豆已收獲而玉米未收獲的時段,在4個待選特征中,NDVI、NDWI與WDRVI指數(shù)組合表現(xiàn)最佳;隨機(jī)森林算法與最大似然算法、支持向量機(jī)算法相比,分類精度更高,其總體分類精度為84.82%,Kappa系數(shù)為77.42%。玉米制圖精度為91.49%,用戶精度為93.48%;大豆制圖精度為91.14%,用戶精度為82.76%。該方法為大區(qū)域農(nóng)作物的分類提供重要參考和借鑒價值。
[Abstract]:It is of great significance for the state to guide agricultural production and formulate agricultural policies to accurately understand the spatial planting and distribution of crops.Aiming at the problem that the growth period of maize and soybean in Heilongjiang Province is close, the spectral characteristics are similar, and it is difficult to distinguish the spectral characteristics, the multitemporal spatial resolution of high resolution GF-1 (GF-1) satellite wide coverage field of view WFV image is taken as the data source.Four characteristics of normalized difference vegetation index NDVI, enhanced vegetation index EVI, wide dynamic range vegetation indexWDRVI, normalized difference water index NDWI) were selected. The random forest classification algorithm was used in combination with field survey sample points.Corn and soybean planting area were extracted from Nenjiang County, Heihe City, Heilongjiang Province.The results showed that the best time to distinguish maize from soybean was from late September to early October, that is, soybean had been harvested but maize had not been harvested, and the combination of NDWI and WDRVI index was the best among the four selected characters.Compared with the maximum likelihood algorithm and the support vector machine algorithm, the stochastic forest algorithm has higher classification accuracy, and its overall classification accuracy is 84.82 and the Kappa coefficient is 77.42.The precision of maize mapping is 91.49, the user accuracy is 93.48, the precision of soybean mapping is 91.14 and the user precision is 82.76.This method provides an important reference and reference value for the classification of crops in large area.
【作者單位】： 中國農(nóng)業(yè)大學(xué)信息與電氣工程學(xué)院;
【基金】：國家自然科學(xué)基金資助(41671418,41471342,41371326)
【分類號】：S513;S565.1;S127
，

本文編號：1697991