基于多级交叉局部二值模式的高光谱图像分类

Hyperspectral image classification based on multi-level cross local binary pattern

高光谱遥感图像的分类一直是比较活跃的研究领域,利用空间纹理特征辅助分类是其中的关键方向之一。针对传统局部二值模式仅考虑中心像元和邻域像元关系,及获得的纹理特征维度过高的问题,提出了利用多级交叉局部二值模式获取空间纹理特征辅助分类的方法。利用局部二值模式算子计算纹理特征时,考虑了邻域像元之间的关系,分别从水平垂直方向和对角方向计算编码值,利用不同尺度窗口生成统计直方图,将其组合获得空间纹理特征。实验表明,此方法能够在合适的维度下获取更有效的纹理特征,在辅助分类过程中进一步提高了分类性能。

The classification of hyperspectral remote sensing images has always been a relatively active research field,and the use of spatial texture features to assist classification is also a key direction.Aiming at the problems that the traditional local binary mode only considers the relationship between the central pixel and the neighboring pixels and the obtained texture feature dimensionality is too high,the method is proposed to obtain spatial texture feature auxiliary classification by using multi-level directional multi-scale local binary mode.When using the local binary mode operator to calculate texture features,the relationship between neighboring pixels is considered,and the coding values is calculated from the horizontal and vertical directions and diagonal directions,and different scale windows are used to generate statistical histograms.Its combination obtains spatial texture characteristics.Experiments show that the method can obtain more effective texture features in appropriate dimensions,and further improve the classification performance in the auxiliary classification process.