针对低信噪比条件下的目标探测问题,提出基于神经网络的动态规划检测前跟踪算法。首先,从原始观测数据中选择训练样本,对神经网络进行权值训练;然后,全部原始观测数据归为两类,将归入杂波集合的数据幅值置零,将归入目标集合的数据进行幅值缩放,改善后续值函数积累效果;最后,提出一种多目标决策航迹回溯算法,对初步回溯的航迹集合进行二次提取,有效降低了航迹回溯过程中的虚假航迹率。仿真结果验证了所提算法的有效性。

In order to address the problem of detecting and tracking low signal-to-noise-ratio targets, a neural network based dynamic programming track-before-detect (NNDP-TBD) algorithm is proposed. First, the training samples are selected from the raw measurements, and the weights of neural network are trained using these samples. Second, all raw measurements are classified into two categories with the clustering analysis. The amplitudes of measurements which belong to the clutter set are set to zero. That of measurements which belong to the target set, however, are increased by a scaling factor. Third, a multiple criteria decision making based backtracking algorithm is proposed to extract again the initial tracks set, which helps to decrease the acceptance probabilities of the false tracks. Simulation results verify the effectiveness of the proposed method.

TN953

国家自然科学基金资助项目