针对典型目标识别问题,提出了一种基于投影寻踪的高光谱目标识别算法。先对高光谱图像进行最小噪声分离变换,计算出本征维度,同时对图像去噪,然后采用信息散度作为投影指标,对投影指标值自适应分割,得到所要提取的波谱曲线,最后用光谱角匹配识别出目标及其位置。高光谱图像验证结果表明,该方法有效地去除了图像噪声,而且能够快速、可靠地提取端元并识别出目标。

In order to solve the problem of typical target recognition, a hyperspectral target recognition algorithm based on projection pursuit is presented. Firstly, the minimum noise fraction(MNF) method is used for the hyperspectral image to calculate the intrinsic dimension as well as image denoising. Then, the information divergence is used as projection index, and the index is adaptively separated to extract the requirable spectral curve. Finally, the target and its location are identified by spectral angle matching method. Through a series of hyperspectral image experiments, the results show that the presented algorithm can not only reduce the image noise more effectively, but also quickly extract endmember and reliably identify the typical target.