光学技术, 2015, 41 (1): 16, 网络出版: 2015-04-14  

一种改进的LPD高光谱图像异常检测算法

An anomaly detection method for hyperspectral image based on improved LPD
作者单位
苏州大学物理科学与技术学院, 江苏 苏州 215006
摘要
在低概率检测(LPD) 算法中,当选取的特征向量数目等于背景地物种类时,算法的检测效果比较理想,然而背景地物的种类数通常不知道,因此难以确定特征向量的数量.针对这一问题,对LPD算法进行了改进:首先用迭代误差分析(IEA)方法提取端元,然后在提取的端元中选择出与背景地物光谱相近的端元,并用它们构成背景矩阵,进而用该矩阵构造出正交投影算子,最后将该投影算子代入到LPD算法中进行目标检测.实验结果表明,该方法可以更有效地抑制背景,降低虚警率,提高检测性能.
Abstract
Only the number of the feature vectors elected is equal to the types of the ground objects,the low probability detection(LPD) has a good effect. The species of background objects usually do not know.It is hard to determine the number of the feature vectors.In order to solve the problem,a novel anomaly detection method based on improved LPD is presented.The iterative error analysis(IEA) algorithm is used to extract the endmembers,and which are similar to the spectrum of background objects are chosen to make up the background matrix.The matrix is applied to build the orthogonal projection operator.The operator is used in the LPD to achieve the target detection.Experimental results show that the proposed method can restrain the background effectively and improve the detection performance obviously.

程凯, 李成金, 赵勋杰. 一种改进的LPD高光谱图像异常检测算法[J]. 光学技术, 2015, 41(1): 16. CHENG Kai, LI Chengjin, ZHAO Xunjie. An anomaly detection method for hyperspectral image based on improved LPD[J]. Optical Technique, 2015, 41(1): 16.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!