基于复杂地面背景序列图像的运动弱小目标检测  

Detection of moving small target based on images in sequence of complicated surface background

作　　者：侯旺[1,2] 欧阳红军[3] 雷志辉[1,2] 刘晓春[1,2] 杨夏[1,2] 

机构地区：[1]图像测量与视觉导航湖南省重点实验室,湖南长沙410073 [2]国防科学技术大学航天与材料工程学院,湖南长沙410073 [3]国防科技大学科研部重点项目处,湖南长沙410073

出　　处：《光学技术》2013年第2期168-173,共6页Optical Technique

摘　　要：分析了复杂地面背景下序列图像检测弱小目标的困难,研究了一种针对场景较复杂的运动弱小目标检测方法。首先将序列图像分割成等间隔的处理单元,并利用最小二乘匹配进行坐标对齐,然后使用Top-Hat算子处理图像并进行CFAR分割,综合考虑处理单元中的每幅图像,对固定背景干扰进行抑制,最后根据运动弱小目标轨迹进行检测。进行了三组序列图像实验并绘制了ROC曲线,证明该方法是一种有效的弱小运动目标检测方法。By analyzing the difficulty of detecting the small targets on complicated surface background in sequence, a detecting method is researched for moving small targets on complicated surface. An integrated detecting method is pro- posed. Firstly, images in sequence are divided into units constituted of the same frames. Its reference frame is unified using LS Matching. Then, images in sequence are processed by Top-Hat and segmented by CFAR. Considering every image in u- nits, the constant background noise is suppressed. At last, the track of small targets is used for detecting. The validity of the proposed method is demonstrated by 3 experiment results and ROC curve.

关 键 词：复杂地面背景 运动弱小目标检测 最小二乘匹配 TOP-HAT 固定背景干扰抑制 RADON变换 

分 类 号：TN911.73[电子电信—通信与信息系统]