基于动态规划的检测前跟踪（DP-TBD）算法研究

发布时间：2018-03-24 13:46

本文选题：弱目标　切入点：检测前跟踪　出处：《大连海事大学》2017年硕士论文

【摘要】：随着雷达技术的迅猛发展,弱目标的检测与跟踪也受到了广泛的关注。与传统的检测方法相比检测前跟踪算法更加高效。当检测与跟踪的目标的信噪比较低时,它能通过对雷达多帧数据的连续处理及沿可能的目标航迹积累能量来同时实现目标的检测与跟踪。其中研究基于动态规划的检测前跟踪算法因为它相对其他算法而言在硬件上易实现,在计算量和存储量上也相对较小。本文对基于动态规划的检测前跟踪算法进行了更加系统的研究,主要内容有:(1)对检测前跟踪算法(TBD)与先检测后跟踪算法(DBT)这两种方法进行的优缺点的比较分析。分析了基于Hough变换的TBD算法、基于三维匹配滤波的TBD算法、基于粒子滤波的方法TBD算法和基于动态规划的TBD算法这四种算法优缺点。搭建了基于动态规划算法的目标运动模型及量测模型。(2)针对传统基于动态规划的检测前跟踪算法的能量扩散效应,提出了三种改进算法。第一种改进算法是在传统动态规划算法中引入新的权重系数对其进行方向加权。针对该算法在低信噪比下检测性能不高的缺点提出了第二种方法,基于三帧积累的检测前跟踪改进算法。该算法通过将二帧递归积累改为三帧递归积累提高了信噪比。为了进一步提高帧间的数据关联性提出了第三种改进算法,基于指数平滑的改进算法。该方法利用预测值和当前帧对应的值的关系更加充分利用了目标的运动性提高了算法的性能。(3)对多目标的检测前跟踪算法进行了研究。首先介绍了基于动态规划的极值法,但是极值法的缺点是无法有效检测目标航迹相交或目标临近的两个目标,针对这一局限性,提出了基于动态规划的剔除法。剔除法是先对目标的强度进行对比,将强度最大的当成目标检测出来,然后将其从观测数据中剔除。如此重复直到检测出所有的目标。
[Abstract]:With the rapid development of radar technology, the detection and tracking of weak targets has received extensive attention. Compared with the traditional detection methods, the pre-detection tracking algorithm is more efficient. When the SNR of the targets detected and tracked is low, It can simultaneously realize the detection and tracking of targets by continuous processing of radar multi-frame data and accumulation of energy along possible target tracks. The pre-detection tracking algorithm based on dynamic programming is studied because it is relative to other algorithms. In terms of hardware, The amount of computation and memory is also relatively small. In this paper, a more systematic study on the pre-detection tracking algorithm based on dynamic programming is carried out. The main contents of this paper are as follows: (1) compare the advantages and disadvantages of the two methods, TBD algorithm based on Hough transform and TBD algorithm based on 3D matched filter. The advantages and disadvantages of TBD algorithm based on particle filter and TBD algorithm based on dynamic programming are presented. The object motion model and measurement model based on dynamic programming algorithm are built. Tracking the energy diffusion effect of the algorithm, Three improved algorithms are proposed. The first one is to introduce new weight coefficients into the traditional dynamic programming algorithm to carry out directional weighting. A second method is proposed to solve the problem that the detection performance of the algorithm is low in low signal-to-noise ratio (SNR). An improved algorithm of pre-detection tracking based on three-frame accumulation is proposed, which improves the SNR by changing the recursive accumulation of two frames into three frames. In order to further improve the data correlation between frames, a third improved algorithm is proposed. An improved algorithm based on exponential smoothing is proposed, which makes full use of the relationship between the prediction value and the corresponding value of the current frame and improves the performance of the algorithm. Firstly, the extremum method based on dynamic programming is introduced. However, the disadvantage of extreme value method is that it can not effectively detect two targets whose track intersects or is close to the target. In view of this limitation, a dynamic programming based culling method is proposed, which is to compare the intensity of the target first. Detect the most intense targets and remove them from the observations. Repeat until all targets are detected.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN953

【相似文献】