| The method for detecting radar target is the premise to determine the radar system,and the simulation technique of radar system is always a crucial step in the advanced radar system research.As an important part of radar system simulation,radar clutter simulation facilitates the verification of the performance and the reliability of radar target detection method in the background of radar clutter.What is more,accurately estimating the parameters of radar clutter is the crucial part of radar signal processing.Therefore,this dissertation mainly studies the method for simulating radar clutter based on the K distribution model,the N-partitioned random multiplicative process model,and the wave spectrum.Also,some methods for estimating the Pareto distributed clutter parameters are improved in this dissertation.1.The shape parameter of the K distributed clutter which is simulated based on the traditional ZMNL and SIRP methods can only take an integral or semi-integral value,which is improved in the dissertation.First,the hitches in the existing methods proposed by Zhu Jieli,E.Conte,and N.D.Wallace are analyzed.All of them generated a Gamma random variable with a fractional shape parameter by the product of a Beta random variable and another Gamma random variable with an integral shape parameter.Second,it is found that the existing methods have a hitch that the histogram of Beta simulated data and its PDF do not coincide well when both the parameters of Beta distribution are not larger than 1.Finally,an improved method for generating a Gamma random variable with arbitrary shape parameter directly by a Beta random variable and another Gamma random variable is proposed.What is more,because the relationship between the autocorrelation coefficients of input and output is for real numbers,only data with symmetrical power spectrum can be simulated by ZMNL.Then,SIRP method is analyzed with mathematical deduction,and it is concluded that the desired shape parameter of the simulated data influences the noise level of its power spectrum.2.The distribution of a multiplier of 2-partitioned random multiplicative model is symmetric about 1/2,which makes the shape of the multifractal function of the simulated data be fixed.Based on the analysis of the construction process of the 2-partitioned random multiplicative model,the N-partitioned random multiplicative model is proposed,as a generalization of the 2-partitioned and it breaks the limitation that the multiplier is symmetric about 1/2.The model is more convenient to simulate data with the desired shape of multifractal function.It is proved theoretically and experimentally that the distribution of the multiplier determines the shape of the multifractal function.3.Sea clutter is simulated by the wave spectrum model.The sea surface can be regarded as a complicated surface spliced by a great many triangular scatterers and the projections of the scatterers on the horizontal plane are right triangles.Therefore the horizontal plane is partitioned by right triangles.Compared with the method based on the distance ring,triangle patches are the equal-area partition.The area of them won't change with the distance,so the details of the sea wave can be considered.Meanwhile,the partitioned method is much easier than that based on the range-Doppler method which needs to calculate double integral.In addition,the sea surface is simulated by the famous wave spectrum method and echo of sea surface is obtained by computing and adding the echoes of all the triangular scatterers,which makes the sea clutter more credible.The experiment shows that the echo of one pulse is in accordance with the real sea clutter and the clutter of the same range fits K distribution best.Thus,the method proposed by this paper can be a way to study the relation between the parameters of the statistical model and that of radar condition.4.The existing methods for estimating the multilook Pareto distribution shape parameter are either non-closed-from or invalid in the small value range.Focused on the problems of the existing methods,some improved methods are proposed:the estimator based on minus moments and the estimators based on?zrln?z??,?z-1ln?z??.Where,?z-1ln?z??estimator is a special case of?zrln?z??estimator mathematically when r=-1.However,r=-1 means that the estimator must be used in a non-coherently integral case.The estimators based on minus moments and based on?z-1ln?z??are closed-form expressions in the whole shape parameter range?0,+??,which overcomes the shortage of estimators based on moments and?zln?z??.Therefore,the computational burden is reduced by avoiding the numerical computation. |
PDF Full Text Request