基于序列图像的小目标检测  

Small target detection based on image sequences

作　　者：贺鹏飞[1] 苏新彦[1] 王鉴[1] 

机构地区：[1]中北大学信息与通信工程学院,山西太原030051

出　　处：《应用光学》2011年第2期272-275,共4页Journal of Applied Optics

基　　金：山西省自然基金(2007012003)

摘　　要：针对序列图像中小目标的运动具有运动的连续性和轨迹的一致性等特征,提出了采用单帧检测和多帧检测相结合的处理方法。对单帧图像进行预处理和图像分割,得到去掉大部分背景和噪声点的图像,并在图像中选取候选目标点进行标记。然后对处理后的序列图像进行N帧叠加。对叠加图中的标定点利用8邻域搜索法进行目标筛选,进而检测出目标。实验结果表明,该方法实用性强,能简单、快速、有效地检测出序列图像中的运动小目标。Based on the moving continuity and trajectory consistency of small targets in the image sequences,a method is put forward for target detection,which combines single frame and multi-frame detection.Firstly,single frame image is preprocessed and segmented,the image is obtained with background and noise removed,and the potential targets are selected and labeled in the image.Secondly,N frames in the processed image sequences are superposed.Finally,eight neighbor search method is used to search the labeled target in the superposed images and detect the target.Experiment results show that the proposed method is simple and fast,which can effectively detect the small moving target in image sequences.

关 键 词：小目标检测 序列图像 中值滤波 图像分割 8邻域搜索 图像叠加 

分 类 号：TN206[电子电信—物理电子学] TP391.41[自动化与计算机技术—计算机应用技术]