光谱快速预选筛分算法研究

Research on Fast Spectral Prescreening Method

利用一组标准正交化的参考向量得到一个可能匹配区域,排除了绝大部分不可能匹配的样本,从而大大提高光谱预选筛分的速度;同时介绍了一种平衡速度与精度的编码替代算法。实验证明:压缩影像的目视效果较好,未产生边缘模糊及条带现象;采用自动目标生成方法(ATGP)从样本集中提取端元并进行光谱角度制图,其效果令人满意。

Using an norm orthonormal set of vectors, on defining a possibility match zone outside most of the exemplars which can not be matched, the prescreening method was accelerated, and a replacement algorithm which balanced the quality with speed was introduced. Our experimentation proveed that： the compress image had a good effect by view, with no fuzz and strip phenomenon： The result was more satisfactory by extracting endmembers from exemplar using the algorithm namely automatic target generation process than using SAM to classify the compress data set.