基于箱粒子概率假设密度滤波的弱目标检测与跟踪算法  

Small targets detection and tracking algorithm using box particle probability hypothesis density filter

作　　者：吴孙勇[1,2] 宁巧娇 蔡如华 孙希延[2] 潘福标 WU Sun-yong;NING Qiao-jiao;CAI Ru-hua;SUN Xi-yan;PAN Fu-biao(School of Mathematics and Computing Science, Guilin University of Electronic Technology, Guilin 541004, China;Guangxi Key Laboratory of Precision Navigation Technology and Application, Guilin 541004, China)

机构地区：[1]桂林电子科技大学数学与计算科学学院,广西桂林541004 [2]广西精密导航技术与应用重点实验室,广西桂林541004

出　　处：《控制与决策》2019年第7期1417-1424,共8页Control and Decision

基　　金：国家自然科学基金项目(61561016,61362005);广西自然科学基金项目(2016GXNSFAA380073,2014GXN SFAA118352,2014GXNSFBA118280);广西密码学与信息安全重点实验室开放基金项目(GCIS201611);广西精密导航技术与应用重点实验室开放基金项目(DH201502);大学生创新创业训练计划项目(201610595036)

摘　　要：针对低信噪比条件下多弱小目标检测前跟踪算法跟踪效率低、计算复杂度高等问题,提出一种基于箱粒子概率假设密度滤波的弱目标检测与跟踪算法。首先,针对由目标的贡献强度和噪声获得的目标强度量测图像,利用均值滤波抑制强度量测图像中的噪声;其次,以不交叉原则挑选出强度值较大区域作为区间量测;最后,利用箱粒子概率假设密度(BOX-PHD)滤波对上述所得的区间量测进行目标跟踪。仿真结果表明,所提出的方法可以提高跟踪性能,且计算效率高。As the track before detect algorithm of small targets has tracking inefficiency and complex computation problems under low signal-to-noise ratio(SNR) condition, a small targets detection and tracking algorithm is proposed by using the box particle probability hypothesis density filter. Firstly, in consideration of the images of the targets’ measured intensity within targets’ contribution intensity and noise, image noise is restrained using the mean filter. Then, the interval measurements are selected from the region of the larger intensity value based on the uncrossed principle. Finally,according to the interval analysis technology, the proposed interval measurements are used to track the targets using the box particle probability hypothesis density(BOX-PHD) filter. Simulation results show that the proposed method can improve the target tracking performance and computation effectively.

关 键 词：箱粒子 区间分析 概率假设密度滤波 弱目标 均值滤波 

分 类 号：TN953[电子电信—信号与信息处理]