论文部分内容阅读
传统的基于贝叶斯模型的谱解混合方法采用了全局端元光谱特征值进行遥感图像的谱解混合,但是不同地理位置以及不同光照条件下,同类地物的端元具有不同光谱特征值,仅仅采用一个全局光谱特征值代替所有地区的光谱特征值降低了遥感图像谱解混合的精度。针对这种情况,提出采用端元局部光谱特征值代替端元全局光谱特征值进行谱解混。具体实施方法是利用相邻地区的同一地物拥有相同的光谱特征值的特点,在提取端元光谱特征值的时候,充分利用空间约束的特点来获取各个混合像元的光谱特征值。实验结果表明,该方法比传统的贝叶斯谱解混合能够得到更高的谱解混合精度。
The traditional Bayesian model-based spectral mixing method uses the spectral values of the global endmember spectrum for the spectral mixture of the remote sensing images. However, the endpoints of the same kind of terrain have different spectral eigenvalues under different geographical locations and different illumination conditions, The mere use of a global spectral eigenvalue instead of the spectral eigenvalues in all regions reduces the accuracy of the spectral mixture of remote sensing images. In view of this situation, it is proposed to use the local spectral characteristic values of endmember instead of the global spectral characteristic values of endmember for spectral unmixing. The specific implementation method is to utilize the same spectral feature value of the same landform in the adjacent area to take full advantage of the characteristics of space constraint to obtain the spectral eigenvalues of each mixed pixel when extracting the spectral characteristic value of the end element. Experimental results show that this method can get higher spectral mixing accuracy than traditional Bayesian spectral mixing.