论文部分内容阅读
目标跟踪中的一个重要问题是在信噪比(SNR)非常低的环境中探测和跟踪目标。以前,使用过包括最大似然的几种方法。本文的主要创新点是把波动目标幅值模型与跟踪等速目标的最大似然方法相结合。与逼真的传感器模型相结合,这种方法可利用同一帧中分辩单元之间以及相邻两帧间的信号相关。 波动幅值模型是反映帧间相关的一阶模型。用卡尔曼滤波器获得幅值估计,由此导出似然函数。以增加计算量为代价,数值最大化技术避免了以前由于假设的和实际的目标速度失配而在“速度滤波”方法中遇到的问题。对不变的已知幅值,导出克莱默—劳下界(CRLB)。即使幅值未知时,估计误差也接近这个CRLB。结果表明,对未知信号幅值的航速探测性能与用正确信号模型获得的性能几乎相同。
An important issue in target tracking is the detection and tracking of targets in very low signal-to-noise (SNR) environments. In the past, several methods including maximum likelihood have been used. The main innovation of this paper is to combine the fluctuating target amplitude model with the maximum likelihood method to track the constant velocity target. In combination with a realistic sensor model, this approach makes use of the signal correlation between the resolution units in the same frame and between two adjacent frames. The fluctuation amplitude model is a first-order model that reflects the inter-frame correlation. The magnitude estimate is obtained with a Kalman filter, from which a likelihood function is derived. At the expense of increased computational effort, the numerical maximization technique avoids the problems previously encountered in the “velocity filtering” approach due to the mismatched assumed and actual target velocities. Krauss-Laulow (CRLB) is derived for a constant, known magnitude. Even if the magnitude is unknown, the estimation error is close to this CRLB. The results show that the speed detection performance for the unknown signal amplitude is almost the same as that obtained with the correct signal model.