中文摘要：

异常检测在高光谱军事应用领域具有重要的意义，它能够在没有光谱库和大气校正的条件下检测出与图像背景存在光谱差异的目标。由于高光谱图像的统计特性复杂，基于统计的异常检测方法在统计模型和参数估计方面存在一定误差，往往不能取得理想的结果。从数据集流形结构的角度，提出了一种基于高维空间超平面结构的高光谱图像异常检测方法。从线性混合模型入手，根据图像数据集合在高维空间中的几何分布的特点，提出了利用超平面的几何特性来检测图像异常的方法。由于算法不使用多元统计模型，具有较强的实用性。最后，利用仿真数据和真实数据对算法的有效性进行检验。

英文摘要：

Anomaly detection plays an important role in hyperspectral military application field, because it can detect the objects whose spectra are different from their surroundings under the conditions of no spectral libraries and atmospheric correction. While the statistic characteristics of hyperspectral image are complicated, some errors exit on the statistic model and parameters estimation so that the anomaly detection methods based on statistic could not acquire the ideal results. From the view of struction of data set,a new approach of anomaly detection based on high dimensionality space hyper-plane in hyperspectral image was presented.Starting from Linear Mixture Modal （LMM）, the geometric distribution characteristic of hyper-plane was used to detect image anomaly,which was based on the geometric distribution characteristic of image＇s data sets in high dimensionality space.The approach is practical for it doesn＇t need multivariate statistic model. Simulation and real datum verify the validity of the approach.