Research On Adaptive Detection Algorithms Of Distributed Targets

With the increasing of radar bandwidth, the size of detection target is always largerthan the distance resoluation cell of wideband radar. Hence, the target of radar does notpresent the traditional point distribution any more but some strong scattering points,which leads the echo of radar target to be distributed in multiple distance cells, i.e., thedetection of target is shown distributed feater. So, the detection of distributed target isresearched deeply in this thesis. Aiming at the Gaussian and compound Gaussian clutterenvironments, the adaptive detection of distributed target is developed as the followingpart:1. The optimal detector and suboptimal detector under the compound Gaussianclutter environments are designed. The suboptimal detector regards the texture ofcompound Gaussian clutter as a deterministic unknown constant and has a simpledesign form and lower calculation. Anylzing performance loss of detector when ignorethe distributed information of texture. At last conclusion is given that the texture ofcoumpound Gaussian clutter has little influence on detection performance of test andcan be ignored.2. Based on one signal observation or multiple observations the adaptive subspacedetector under the fist-order and second-order Gaussian models in partiallyhomogeneous environments is researched. Aiming at different kinds of observations wepresent the design methods of test with the noise covariance maxtrix is known orunknown. The closed expressions of probability of false-alarm (PFA) and probability ofdetection (PD) of the tests under the first-Gaussian model&second-Gaussian modelwithout knowing the noise covariance maxtrix are got, respectively. Moreover, theexpressions of PFA of the adavptive subspace detector have a constant false alarm rateto scaling factor and noise covariance maxtrix. Finally the simulation result is given.3. Establishing the noise of distributed target detection as an autoregressive modelwith lacking of secondary data, considering two types of noise eviroments, that is,homogeneous environment and heterogeneous environment. The tests have the sameasymptotic performance of the generalised likelihood ratio test (GLRT) in the two kinds of noise eviroments. Simulation result shows that the tests have a good detectionperformance without secondary data in homogeneous environment and heterogeneousenvironment.4. Proposing a full-scaling modulus maxmiun denosing algorithm of distributedtargets according to the difference of Lipschitz exponent between noise and signal,which can be used to search the modulus maximum points and separate noisy signal ona full scale. Meanwhile, the singularities can be reserved at the most extent by adaptivethreshold and preprocessing to modulus maximum sequences. The reconstructionalgorithm is based on polynomial interpolation functions. The result has shown thatcompared with the classical Mallat algorithm the proposed algorithm is increased byover10%in the SNR of the denoised signal, the RMSE is reduced by35%on average,and the processing time is reduced by88%.