基于端元变化的两种混合像元分解算法比较研究

发布时间：2019-02-21 20:39

【摘要】：光谱混合分析对提高遥感影像分类具有重要意义,其中端元变化处理是提高解混精度的关键。目前,许多算法被用来解决端元变化,但仍存在一些问题有待解决,如算法运行效率慢、忽略端元的高阶交互、像元空间邻城信息缺失。结合IDL和MATLAB混合编程,利用确定性模型中的交替最小角度法和统计性模型中考虑高阶项的非线性算法对Hyperion影像进行端元变化解混,再利用概率松弛标记法对像元补充空间邻域信息。试验结果表明:当某种地物类别所占面积较大时,确定性与统计性模型都能获得较高的解混精度;当地物类别所占面积较小时,确定性模型的解混精度高于统计性模型;补充像元空间邻域信息对解混结果有很好的校正。
[Abstract]:Spectral mixing analysis plays an important role in improving the classification of remote sensing images. End-component change processing is the key to improve the resolution of remote sensing images. At present, many algorithms are used to solve the endelement change, but there are still some problems to be solved, such as the low efficiency of the algorithm, the neglect of the higher-order interaction of the endelements, and the lack of information in the pixel space adjacent to the city. Combined with IDL and MATLAB, the alternating minimum angle method in the deterministic model and the nonlinear algorithm considering higher order terms in the statistical model are used to unmix the end element variation of the Hyperion image. Then the probabilistic relaxation marking method is used to supplement the spatial neighborhood information to the pixel. The experimental results show that both deterministic and statistical models can obtain higher decontamination accuracy when the area of a certain object category is larger, and the precision of deterministic model is higher than that of statistical model when the area occupied by local object category is small. The neighborhood information of complementary pixel space is well corrected.
【作者单位】：西华师范大学国土资源学院;
【基金】：国家自然科学基金面上项目(编号:41671220) 西华师范大学博士科研启动基金项目(编号:412546、412547) 四川省教育厅自然科学基金重点项目(编号:15ZA150)
【分类号】：TP751

【相似文献】