CAJ | 학술논문

针对高光谱图像背景复杂导致高光谱图像异常检测效果下降的问题，提出了一种新的基于抑制背景的高光谱图像异常检测方法。该方法首先使用主成分分析法抑制高光谱图像中的背景信息，得到背景抑制后的图像，然后再使用基于核的RX算法（ KRX算法）异常检测，最后将检测结果图进行阈值分割，得到一幅二值图像。最后使用ROC曲线对检测结果进行评价，通过与RX、KRX算法对比，证明本文方法得到的结果具有较高的检测率和较低的虚警率，充分说明文中方法的有效性。
침대고광보도상배경복잡도치고광보도상이상검측효과하강적문제，제출료일충신적기우억제배경적고광보도상이상검측방법。해방법수선사용주성분분석법억제고광보도상중적배경신식，득도배경억제후적도상，연후재사용기우핵적RX산법（ KRX산법）이상검측，최후장검측결과도진행역치분할，득도일폭이치도상。최후사용ROC곡선대검측결과진행평개，통과여RX、KRX산법대비，증명본문방법득도적결과구유교고적검측솔화교저적허경솔，충분설명문중방법적유효성。
In order to solve the problem that the complex background will reduce the effect of anomaly decetion, a novel anomaly detection algorithm was proposed.The background of the hyperspectral image is suppressed by prin-cipal component analysis (PCA), and the image after processing was detected by Kernel RX (KRX) algorithm.A binary image can be obtained after threshold segmentation.The higher detection result and the lower false alarm can be gotten by this method, which can prove its advantages by receiver operating characteristic curve ROC) compared with RX and KRX algorithm.