基于改进最小噪声分离变换的异常检测算法

为了降低噪声对高光谱异常检测结果的影响以及提高异常检测率,提出了一种基于改进最小噪声分离(MNF)变换的新型高光谱异常检测算法。首先对传统的MNF变换进行改进,采用加权邻域均值法对噪声矩阵进行估计,对邻域内每一个像元给予一个特定的权值,提高背景像元在邻域矩阵中的比例,进而抑制噪声像元的比例,通过差值计算提取噪声信息,然后应用改进的MNF变换对高光谱图像进行降维去噪处理,最后,将获取的低维去噪图像利用异常检测算法进行检测,并用真实的AVIRIS数据进行了测试。结果表明,该算法有更好的降维去噪效果,提高了异常检测率。

Anomaly detection method based on improved minimum noise fraction transformation

In order to reduce the influence of noise on the detection results of hyperspectral anomaly detection and improve the rate of anomaly detection,a new anomaly detection process based on improved minimum noise fraction (MNF) transformation was proposed. Firstly, to improve the traditional MNF transform, the weighted neighborhood averaging method was used to estimate the noise matrix,a specific weight was given to each pixel of the neighbor matrix for increasing the portion of background pixels and suppressing the noise pixels in the sample matrix. It was an effective way to extract noise information by calculating the difference. Secondly, improved MNF transform was used to reduce the dimension of hyperspectral image data and to separate the noise from signals effectively.Finally, anomaly detection algorithm was implemented on low-dimensional denoised data. After actual test of AVIRIS data, the results show that the improved algorithm has better effect of reducing the dimension and separating the noise, and the rate of anomaly detection is improved significantly.