论文部分内容阅读
针对无约束最小二乘混合像元分解算法提取地物端元丰度出现的局限性问题,通过野外实地采集的地物光谱数据建立研究区典型的地物波谱库,以Landsat OLI影像作为主要数据源,在经过Gram-Schmidt(GS)影像融合的基础上,利用纯净像元指数(PPI)及基于几何顶点的端元提取技术提取研究区典型地物端元,最后通过完全约束的最小二乘混合像元分解算法完成对研究区典型地物端元丰度的提取。结果较好地解决了无约束最小二乘混合像元分解算法提取的端元丰度信息出现负值的情况,并且提高了典型地物丰度信息提取的精度。完全约束最小二乘混合像元分解算法的RMSE误差均控制在0.174 913左右,在很大程度上提高了混合像元分解精度及实用性。
In view of the limitations of the unconstrained least squares mixed pixel decomposition algorithm for extracting terminal abundances, a typical object spectral database of the study area was established based on the spectral data of the ground objects collected in the field. Landsat OLI images were taken as the main data Based on the image fusion of Gram-Schmidt (GS), we extracted the end-points of the typical landforms in the study area using the pure pixel index (PPI) and the end-point extraction based on geometric vertices. Finally, Mixed pixel decomposition algorithm to complete the typical terrain end of the study area abundance extraction. The result shows that the terminal element abundance information extracted by the unconstrained least squares mixed pixel decomposition algorithm has a negative value, and the precision of extracting the abundance information of typical features is improved. The RMSE error of the fully constrained least squares mixed pixel decomposition algorithm is controlled at about 0.174 913, which improves the resolution and practicability of mixed pixel decomposition to a great extent.