CAJ | 학술논문

高光谱图像背景建模的好坏会直接影响目标检测的效果。为了建立适宜目标检测算法的背景，提出了一种基于图像差分的自适应背景建模方法。首先对地物光谱特征点所在波段图像进行差分，之后根据差分结果的正负关系进行0、1编码，再对图像进行初步分类后进行闭运算处理，最后建立背景用于目标检测。通过异常检测和目标识别的实验，证明了该方法能够大幅提高目标的检测率，减少误警率，并提高了运算速度。
고광보도상배경건모적호배회직접영향목표검측적효과。위료건립괄의목표검측산법적배경，제출료일충기우도상차분적자괄응배경건모방법。수선대지물광보특정점소재파단도상진행차분，지후근거차분결과적정부관계진행0、1편마，재대도상진행초보분류후진행폐운산처리，최후건립배경용우목표검측。통과이상검측화목표식별적실험，증명료해방법능구대폭제고목표적검측솔，감소오경솔，병제고료운산속도。
The effectiveness of hyper-spectral images target detection is affected by the quality of background modeling directly. In order to model best background, A new method of adaptive background modeling based on image difference was presented. First, the bands where the ground object spectral' s feature points locate were differenced. And the result of image difference was encoded into binary format. Then close operation was conducted after raw image classification. At last, background was modeled to detect targets. Anomaly detection and target recognition experiments show that this method can enhance the target detection rate remarkably, decrease mistake warning rate and improve operation speed.