系统工程与电子技术 ›› 2019, Vol. 41 ›› Issue (11): 2479-2487.doi: 10.3969/j.issn.1001-506X.2019.11.11

• 传感器与信号处理 • 上一篇    下一篇

基于隐马尔可夫模型的动态规划检测前跟踪算法

张袁鹏, 郑岱堃, 李昕哲, 孙永健   

  1. 空军预警学院一系, 湖北 武汉 430019
  • 出版日期:2019-10-30 发布日期:2019-11-05

Dynamic programming track-Before-detect algorithm based on hidden Markov model

ZHANG Yuanpeng, ZHENG Daikun, LI Xinzhe, SUN Yongjian   

  1. The First Department, Air Force Early Warning Academy, Wuhan 430019, China
  • Online:2019-10-30 Published:2019-11-05

摘要: 传统的动态规划检测前跟踪(dynamic programming track-before-detect, DP-TBD)算法能有效实现对匀速直线运动目标的检测跟踪,但其忽略了目标帧间状态转移概率,因此在对机动目标进行检测跟踪时容易受噪声干扰,产生错误的状态关联。对此提出了一种基于隐马尔可夫模型的DP-TBD算法。该算法利用隐马尔可夫模型对目标的运动过程建模,用一系列隐状态表示目标转弯速率并利用隐马尔可夫模型的隐状态估计理论实现对转弯速率的估计和预测,进而得到当前目标状态的预测值,根据此预测状态与下一时刻回波数据分辨单元间的距离来计算转移概率。然后将转移概率应用于DP-TBD算法的能量积累过程中以提高检测跟踪性能。仿真实验基于机动目标,给出了所提算法的检测跟踪性能,并与传统的DP-TBD算法、方向加权DP-TBD算法以及线性最小二乘DP-TBD算法进行了分析比较,验证了该算法的有效性。

关键词: 检测前跟踪, 动态规划, 隐马尔可夫模型, 状态预测, 状态转移概率

Abstract: The traditional dynamic programming track-before-detect(DP-TBD)algorithm can effectively detect and track targets of uniform linear motion. However, the ignorance of the target’s state transition probability among successive flames results in an incorrect state association and it is easy to be disturbed by the noise when detecting and tracking maneuvering targets. This paper proposes a DP-TBD algorithm based on the hidden Markov model (HMM). The algorithm uses the HMM to describe the motion of the target. A series of hidden states are used to represent the turning rate of the target, which are also estimated and predicted by the hidden state estimation theory of the HMM to the turning rate. The predicted value of the current target state will be obtained at the same time. According to the distance between the predicted state and the next echo data, we calculate the state transition probability. Then this probability is applied to the energy accumulation process of the DP-TBD algorithm to improve the detection and tracking performance. Based on the maneuvering target,the simulation gives the detection and tracking performance of the proposed algorithm, compares the proposed algorithm with the traditional DP-TBD algorithm, the direction-weighted DP-TBD algorithm and the linear least squares DP-TBD algorithm,and the effectiveness of the proposed algorithm is verified.


Key words: track-before-detect (TBD), dynamic programming (DP), hidden Markov model (HMM), state prediction, state transition probability