Research On Knowledge Aided Track-before-detect Algorithm

In modern radar systems, the growing complexity of the environment and target makes the detection of dim (low signal-to-noise/clutter ratio) targets more challenging.Compared with conventional detect-before-track (DBT) technique, track-before-detect(TBD) technique can improve the detection performance effectively by jointly processing several consecutive data frames. Dynamic programming (DP) is an efficient way to implement TBD, since it is easy to implement and of good performance, it attracts more and more attention. In the case where the background clutter is uniformly and Gaussian distributed, while the target is a point one and non-fluctuating, dynamic programming based TBD (DP-TBD) can achieve good performance. However, in the case where the background clutter is non-uniformly distributed, or the background clutter is non-Gaussian distributed, or the target is an extended one, DP-TBD suffers from significant performance loss. This is due to the fact that the implementation of DP-TBD relies on the measurement amplitude, but the measurement amplitude in these cases can not reflect the difference between the target and background clutter in signal well. The challenging aforementioned can be overcome resorting to the knowledge aided technique, the detection performance can be enhanced through the use of the knowledge of environment and target. Thus, in this thesis, the knowledge aided technique and DP-TBD are combined, and the knowledge-aided DP-TBD algorithm is studied to improve the detection performance of dim target. The main contributions of this thesis are given as below:1. For non-uniformly distributed background clutter, fixed threshold based and clutter partition knowledge aided DP-TBD algorithm, as well as adaptive threshold based and clutter partition knowledge aided DP-TBD algorithm, is proposed, with which the false tracks produced in the partition with higher clutter power can be reduced.In addition, the detection performance of the target moving within the partition with lower clutter power can be enhanced.2. For two typical non-Gaussian distributed clutter (K and GO) and three typical target amplitude fluctuation models (Swerling 0, 1 and 3), the amplitude knowledge aided DP-TBD algorithm is proposed. By using amplitude information in the DP integration, the performance loss produced by the "heavy-tailed" clutter measurements(clutter measurements with large amplitude) can be reduced.3. To detect extended target (three kinds of extended target are considered, i.e.,Swerling 0, 1 and 3) in Gaussian distributed clutter, measurement amplitude based and extended target feature knowledge aided DP-TBD algorithm is proposed. By using extended target feature information in the DP integration, the detection performance of the extended target in Gaussian distributed clutter is enhanced. In addition, complex measurement based and extended target feature knowledge aided DP-TBD algorithm is proposed. By using the extended target feature information in the DP integration, the background clutter is further suppressed and the signal-to-clutter ratio is enhanced since the phase information is retained as well as the measurement amplitude.4. To detect extended target in compound-Gaussian distributed clutter,non-coherent integration based and extended target feature knowledge aided DP-TBD algorithm is proposed. By implementing non-coherent integration on the measurement amplitude inter-frame using the extended target feature information, the utilization of the energy of the extended target is improved. Moreover, generalized likelihood ratio test (GLRT) and extended target feature knowledge aided DP-TBD algorithm is proposed. By implementing non-coherent integration on the GLRT statistics of the measurement amplitude inter-frame using the extended target feature information, the performance loss produced by the clutter measurement with large amplitude can be reduced.