Research On Wide Band Radar Extended Target Track-Before-Detect

Target detection and tracking is a key function of modern military and civilian perception information systems.Its basic task is to use the noisy data acquired by sensors to estimate the number of targets and their corresponding states in the perceived scene.The idea of classical perceptual information processing is to separate this task into two stage.Firstly,the target detection is realized by using the constant false alarm detector,and then the detection result is input into the tracking filter to achieve the target state estimation.This kind of detect-before-track idea has achieved rich research results under high SNR conditions,but it will lead to missed detection of weak targets under low SNR conditions.The track-before-detect idea directly uses the sensor input video signal data,which utilizes the non-coherent energy accumulation on the potential target track to better detect and track the weak target.Currently,most of the track-before-detect filters use the point target hypothesis.The resolution of the sensor is increasing day by day and a target often occupies multiple resolution cells.Therefore,this research takes wide band radar target detection and tracking as the application background.Under the condition that it is difficult to obtain the prior information of extended target amplitude in practical application,this research focuses on solving the problem of detection and tracking of single extended target and time varying number of extended target.The subject utilizes the advantages of finite set statistics theory in time-varying multiple target modeling to complete the construction of single or multiple extended target kinematic models,and completes the tasks on estimating the unknown extended target amplitude information based on the extended target measurement accumulated data on the potential track.Both through the recursive Bayesian filtering tool,the joint detection and tracking of single or multiple extended targets is finally realized.And the extended target amplitude distribution information of the recursive estimation can be applied to the future target rough classification.In the third chapter of the thesis,a new single-target Bernoulli track-before-detect filter is established for the problem that the target prior amplitude distribution is unknown under the point target hypothesis.The filter uses the multi-frame radar video measurement data on the target potential track to recursively estimate the target average echo amplitude and apply it to the filter's recursive update step to achieve weak target detection and tracking.The experimental results show that the proposed method can better realize the detection and tracking under the scene condition despite the detection delay,and gradually converges to the optimal filter whose prior information is known.Under the condition of single extended target hypothesis,this thesis also establishes a single extended target Bernoulli track-before-detect filter for the problem that the number of resolution cells and the amplitude distribution information of the extended target are unknown.The overall idea is similar to the point target hypothesis.However,in view of the problem that there may be extraneous noise samples in the multi-frame radar video measurement data accumulated on the potential track of the extended target,the mathematical tool of principal component analysis is introduced to achieve robust estimation of the extended target amplitude distribution information.The filter overcomes the divergence problem existing in the Swerling 1 targets.The simulation results show that the weak target detection and tracking under the condition of unknown extended target length and amplitude distribution can be effectively realized,and is asymptotically convergent to the optimal filter when the information is known a prior.The fourth chapter of the thesis further expands the single extended target trackbefore-detect idea of unknown amplitude prior information in Chapter 3 to the timevarying number multiple target scenario.Based on the advantages of finite set statistics theory,a multi-Bernoulli motion model is established with time-varying number.The unknown target prior information is estimated by the extended target measurement data on the potential track,which is used to calculate extend target observation likelihood function.The joint detection and tracking of multiple extended targets is then realized.The simulation results show that the filter can achieve multiple extended target joint detection and tracking of unknown amplitude prior information.