基于改进多特征融合均值位移算法的红外舰船目标跟踪  

Infrared ship tracking based on improved multi-features fusion based mean-shift

作　　者：赵菲[1] 卢焕章[1] 张志勇[1] 

机构地区：[1]国防科学技术大学ATR国防科技重点实验室,湖南长沙410073

出　　处：《系统工程与电子技术》2014年第2期205-213,共9页Systems Engineering and Electronics

基　　金：国家高技术研究发展计划(863计划)(2009AA8100701)资助课题

摘　　要：针对红外海面背景杂波干扰严重,多特征融合的均值位移算法无法准确跟踪目标的问题,对原算法进行了改进,提出一种基于多特征融合均值位移跟踪的新框架。在多特征融合均值位移跟踪的基础上,所提算法通过加入对目标区域的迭代分割并提取区域形心,实现了舰船目标的准确再定位;在局部分割的基础上,提出了对多特征模板尺寸和目标模型的更新方案,实现了对红外海面背景下舰船目标的准确、鲁棒跟踪。利用实际红外图像序列对算法性能进行验证,实验结果表明,所提算法能够更准确、有效地跟踪图像序列中的舰船目标。Because of the heavily cluttered infrared sea background, the original multi-features fusion based mean-shift（MFMS） can not track the target accurately. The original MFMS algorithm is improved, and a new MFMS tracking framework is proposed. Based on the tracking results o{ MFMS, iterative seg- mentation in local area is performed and the center of region is extracted, so the target can be located accu- rately. Based on the segmentation result, the update scheme for template bandwidth and target model is pro- posed. The target can be tracked accurately and robustly. The proposed algorithm is applied to real infrared image sequences. As experimental results demonstrated, the proposed algorithm can track the ship target more accurately and effectively.

关 键 词：均值位移 多特征 海面红外背景 舰船跟踪 

分 类 号：TN911.73[电子电信—通信与信息系统]