CAJ | 학술논문

针对天空背景红外图像中弱小目标检测的难题，分析了红外目标检测的模型，提出了基于稀疏环决策的目标检测算法。利用数学形态学滤波目标增强方法对图像进行背景抑制，而后采用恒虚警检测方法对滤波后图像进行自适应分割，从而获得候选目标点，然后计算各个候选目标点的局部自相似性描述子，对自相似性描述子归一化、分块之后得到稀疏环表示，利用相应的判断准则可以判别目标点与虚警点。实验结果表明，该算法应用于复杂云层背景弱小红外目标图像能够得到较理想的结果，与移动管道滤波方法相比，能有效区别目标点与固定云层杂波干扰，并且虚警率低，易于实现。
침대천공배경홍외도상중약소목표검측적난제，분석료홍외목표검측적모형，제출료기우희소배결책적목표검측산법。이용수학형태학려파목표증강방법대도상진행배경억제，이후채용항허경검측방법대려파후도상진행자괄응분할，종이획득후선목표점，연후계산각개후선목표점적국부자상사성묘술자，대자상사성묘술자귀일화、분괴지후득도희소배표시，이용상응적판단준칙가이판별목표점여허경점。실험결과표명，해산법응용우복잡운층배경약소홍외목표도상능구득도교이상적결과，여이동관도려파방법상비，능유효구별목표점여고정운층잡파간우，병차허경솔저，역우실현。
Aiming at the problem of dim target detection in infrared images with sky background,the detection model of infrared targets is analyzed and a detection algorithm was put forward based on the sparse ring decision.The morphological filtering target enhancement method was used for background suppression,then the Constant False Alarm Rate ( CFAR ) detection method was adopted for image adaptive segmentation to get candidate target points.The Local Self-Similarity (LSS) descriptors of candidate target points were calculated out,and sparse ring was obtained by normalizing the LSS descriptors and partitioning .By means of appropriate criterion,the target point and false alarm points can be distinguished .Experiments show that:1) The algorithm can get ideal results when applied to infrared images with dim and small target against clutter cloud background;and 2) Comparing with moving pipeline filter algorithm,it has lower false alarm rate and is easier for implementation .