基于谐波分析和线性混合模型的河北平原区土地覆被分类研究

Land Cover Classification of Hebei Plain Based on Harmonics Analysis and Linear Mixture Model

土地覆被分类是研究土地利用/覆被变化的基础数据和关键环节。以16d合成MODIS-EVI时间序列数据为主要数据源,采用谐波分析方法分析不同土地覆被类型的季节性变化规律和物候特征差异,引入谐波特征值构建线性混合模型,提取不同端元的丰度值。从土地覆被类型较齐全、谐波特征具有代表性的石家庄地区高空间分辨率影像上选择训练样本,确定MODIS纯净像元和混合像元的划分阈值,对河北平原区进行土地覆被分类制图。结果表明,与河北省县级土地调查统计数据对比,一年两熟耕地、一年一熟耕地、园地及有林地、自然陆地表面的总量精度分别为90.19%、86.17%、85.96%和77.82%,平均总量精度为85.03%;与石家庄地区9个县(市)一年两熟耕地和一年一熟耕地基于TM的分类结果对比,平均面积相对误差分别为10.25%、13.98%。受粗空间分辨率和合成周期、水热条件以及种植模式破碎化限制,混合像元主要集中在河北平原中东部地区,一年两熟耕地、一年一熟耕地、园地及有林地混合面积比重较大。

In this paper, the time series data of 16-days composite MODIS Enhanced Vegetation Index （EVI）were used to exam- ine the land use/cover classification in Hebei Plain. The harmonic analysis of time series （HANTS）algorithm based on Fourier transformation was applied to predigest and compress the MODIS-EVI time series images, at the same time, the phenological features of two crops a year, one crop a year, garden woodland and natural land were extracted. The first to third eigenvalues were used to establish the linear mixture model and calculate the abundance of different endmembers. Shijiazhuang district, which had all types and characteristics of land cover and eigenvalue,was used to set the thresholds of the pure and mixed pixel based on the correlation between MODIS and Landsat images. The land cover mapping was completed by these thresholds. The result indicated that the overall accuracy of two crops a year, one crop a year, garden woodland and natural land was 90. 19 %, 86.17 %, 85.96 %, 77. 82% with the average was 85. 03% when compared with official statistics at county-level. The average area error of two crops a year and one crop a year in Shijiazhuang were 10. 25% and 13. 98%. Mixed pixels are mainly concen- trated in middle east of Hebei Plain and larger proportion is two crops a year,one crop a year,garden woodland. The error was caused mostly by the coarse resolution and composite period of MODIS, the water and heat conditions and the great spatial vari- ability.