基于多尺度分形与均值漂移的红外海面舰船目标分割  

Infrared ship object segment based on MFFK and MeanShift filtering

作　　者：邹常文[1] 刘先志[1] 戴军[1] 严发宝[2] 

机构地区：[1]西南技术物理研究所,四川成都610041 [2]西南自动化研究所,四川绵阳621000

出　　处：《激光与红外》2010年第9期1023-1026,共4页Laser & Infrared

基　　金：预研项目资助

摘　　要：基于分形检测理论,结合均值漂移滤波算法做了一些红外舰船目标分割提取实验。利用多尺度分形特征MFFK检测舰船目标得到感兴趣区域ROI,去除了感兴趣区域外的大面积背景干扰;对感兴趣区域的分形特征图进行二值化处理再次减少了感兴趣区域内的背景干扰,经过两步去除干扰得到二值目标区域。对二值目标区域进行均值漂移滤波得到真实目标分布点集,最后运用形态学滤波得到舰船目标分割结果。实验结果表明,提出的算法能够提取出舰船目标。The paper makes some experiments on infrared image segmentation of ship object based on the fractal theory and MeanShift filtering.Using MFFK to pick out the region of interest（ROI） can remove large area background noise out of the ROI.Then,getting the binary mask of ROI will suppress the noise of the ROI further.Most of the noises maybe excluded via these two procedures.So using MeanShift filtering can get the distribution points of the target.At last,morphological filtering is adopted to obtain the object ship image.The experiment results show that the proposed algorithm can extract the ship object well.

关 键 词：红外舰船目标 分割 多尺度分形特征 均值漂移 形态学滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]