Research On Coast-to-ocean Inverse Synthetic Aperture Radar Target Detection Technology

Radar always works in the complex environment, such as strong clutter, humaninterference and so on, which brought great challenge to target detection. Thetraditional threshold detection is far from the desired effects, especially for thedetection on weak targets. In order to find the target more effectively and reliably,in the complex environment, track before detect technology is one important meansof processing weak target detection problem. TBD is a technique, where tracks anddetections are produced directly on the basis of raw measurements, power or I-Qdata, without intermediate processing making. The advantage over traditionaltracking and detection method is that the useful measurement data is accumulatedover time, so weaken clutter signal at the same time highlights the targetinformation, this leads to the improvement on the signal-noise-ratio and targetdetection performance.This paper to solving the nonlinear TBD problem, adopted the dynamicprogramming and particle filter algorithms, which improved the weak targetdetection and tracking performance. The main contents are as follows:Firstly, creating the TBD processing model based on the hypothesis of uniformmotion in a straight-line and point target spread. Introduced the basic principles ofdynamic programming algorithm, on this basis establish the algorithm’s frameworkand DP-TBD implementation process. Verified and analysis the performance of thealgorithm by simulation. Against the large computation of the traditional DP-TBD,in the chapter2introduced an improved DP-TBD processing method, and analysisthe different pre-threshold influence on the detection performance.Secondly, describes the basic principles of the particle filter theory, and givingdetailed analysis of the composition of the particle filter technique, such assequential importance sample、resampling technique and so on. In the end, verifiedthe tracking performance of the particle filter based on the pure azimuth observationexample. And analysis the algorithm’s error performance based on the root meansquare error.Finally, on foundation of the basic particle filter theory, accomplished the modelestablishment of TBD theory based on particle filter by adding discrete variableswhich represents the target existence. This TBD technique allows the user toimplement any optimal detection in terms of the weights of particles and particles.In other words, allows approximate the distribution of posterior probability density and the detection probability by mixed state estimation, and by that way implementthe target tracking and detection. In the last Section, proved the algorithm’sdetection performance based on the radar sensor observation model. Analysis thedifferent SNR and number of particles influence on the PF-TBD.