基于时间信息加权的ViSAR目标阴影跟踪算法  

Shadow tracking algorithm for ViSAR target based on time information weighting

作　　者：黄艳堃 邓玉辉 孙光才[1] 邢孟道[1] HUANG Yankun;DENG Yuhui;SUN Guangcai;XING Mengdao(National Laboratory of Radar Signal Processing,Xidian University,Xi’an 710071,China)

机构地区：[1]西安电子科技大学雷达信号处理全国重点实验室,陕西西安710071

出　　处：《系统工程与电子技术》2023年第10期3065-3075,共11页Systems Engineering and Electronics

基　　金：国家自然科学基金杰出青年基金(61825105)资助课题。

摘　　要：基于目标阴影的跟踪技术是视频合成孔径雷达(video synthetic aperture radar,ViSAR)目标探测的重要手段,但ViSAR数据存在目标特征不明显且随时间不规则变化、相干斑噪声干扰强等问题,使得ViSAR目标阴影跟踪精度较低。为此,提出了一种鲁棒的基于时间信息加权的ViSAR目标阴影跟踪算法。针对目标特征不明显且随时间不规则变化的问题,将尺度自适应均值偏移(adaptive scale mean shift,ASMS)跟踪算法引入到ViSAR目标阴影跟踪中,同时在ASMS算法的背景比例加权(background ratio weighted,BRW)技术中添加历史帧的特征,并对尺度正则项进行时间信息加权修正,来对目标特征进行整合。针对相干斑噪声干扰强的问题,对ASMS算法加入局部中值滤波操作的预处理步骤,在不增加计算量的同时平滑了噪声。在对两类ViSAR数据集上不同尺寸、不同运动状态的目标阴影的跟踪实验结果表明,与现有高性能跟踪算法相比,所提算法在保证了实时性的基础上提高了跟踪精度,且不需要额外的训练样本,具备较好的工程应用价值。Target shadow-based tracking technology is an important method for target detection in video synthetic aperture radar(ViSAR).However,there are some problems in ViSAR data,e.g.,unclear target characteristics,irregular change of target characteristics with time,strong speckle noise interference,etc.,which makes ViSAR target shadow tracking less accurate.Therefore,a robust ViSAR target shadow tracking algorithm based on time information weighting is proposed in this paper.Aiming at the problem that the target characteristics are not obvious and change irregularly with time,the adaptive scale mean shift(ASMS)tracking algorithm is introduced to ViSAR target shadow tracking.Meanwhile,the characteristics of historical frames are added to the background ratio weighted(BRW)technology of the ASMS algorithm.Target features are integrated with the time information weighted correction to the scale regularity term.Aiming at the strong interference of speckle noise,the preprocessing step of local median filtering operation is added to the ASMS algorithm to smooth the noise without increasing the amount of calculation.The experimental results of tracking target shadows with different sizes and different motion states on two kinds of ViSAR dataset show that,compared with the existing high-performance tracking algorithms,the proposed algorithm improves the tracking accuracy while ensuring real-time performance,and does not require additional training samples,which has good engineering application value.

关 键 词：视频合成孔径雷达 目标跟踪 尺度自适应均值偏移 时间信息加权 中值滤波 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]