论文部分内容阅读
针对风云3数据的特点,本文将EVI生长曲线引入多光谱混合像元的分解。首先,利用Landsat8 OLI影像,采用支持向量机的分类方法,提取研究区域的耕地信息,利用该信息对风云MERSI数据进行掩膜处理,获得研究区域的耕地影像。接着,利用MERSI时序影像,计算像元EVI值,通过SG滤波,构建农作物(端元)和混合像元的EVI生长曲线。通过实地调查,获取研究区的农作物端元,尤其对主要的农作物玉米,在空间上均匀选取了14个端元。然后,采用传统的方法,将14种玉米端元生长曲线分别与其它端元组合,进行混合像元分解。发现分解的效果差异很大,提取的玉米种植面积从191.90 km2到574.83 km2不等。为提高分解精度,借用光谱匹配(光谱夹角最小)的方法(用生长曲线代替光谱曲线)自适应选择与混合像元EVI曲线最相似的玉米端元作为组合端元,进行混合像元分解。结果得到玉米的种植面积为589.95 km2,比传统方法的最好(相对)精度提高了2%。
According to the characteristics of FY-3 data, we introduce the EVI growth curve into decomposition of multi-spectral hybrid pixels. Firstly, using Landsat8 OLI images, the classification method of support vector machines was used to extract the cultivated land information of the study area, and the cloud cover MERSI data was masked to obtain the cultivated land image of the study area. Then, by using the MERSI time-series images, the pixel EVI values are calculated, and the SGI filter is used to construct the EVI growth curves of crops (end units) and mixed pixels. Through the field survey, the end-crop of the crop in the study area was obtained, especially for the main crop maize, 14 end-points were spatially evenly selected. Then, using the traditional method, 14 corn endmember growth curves were separately combined with other end units for mixed pixel decomposition. The effect of decomposition was found to vary widely, with corn acreage ranging from 191.90 km2 to 574.83 km2. In order to improve the decomposition accuracy, the mixed-pixel decomposition was performed by using the method of matching spectra (the angle of the spectrum is the smallest) (using the growth curve instead of the spectral curve) to adaptively select the corn endmember that is most similar to the mixed pixel EVI curve as the combined endmember. As a result, the area planted with corn was 589.95 km2, a 2% improvement over the best (relative) accuracy of the traditional method.