Radar Scanning Beam Super-resolution Based On Statistical Optimization

Spatial resolution is an important index that affects radars’ detection performance, including range resolution and angular resolution. The angular resolution is mainly limited by the antenna size,and is ususally improved by synthetic aperture radar(SAR) and Doppler beam sharpening(DBS) techniques depending on the relative motion between the radar and targets that can produce angle change or Doppler frequency variation. However, because of restriction of the mechanism, they can not meet many application requirements for warning radar, surveillance radar and airborne scanning radar system etc. Therefore, researching and exploring new techniques of improving radar angular resolution has important significance.In this dissertation, radar scanning beam super-resolution based on statistical optimization is researched in order to improve radar angular resoluiotn. Following works mainly include echo modeling, super-resolution algorithms and iterative stopping problems etc. Main works are as follows:1.On the basis of the relative geometric relationship between radar and targets, according to characteristics of radar beam scanning, the temporal and spatial relationship between antenna beam scanning and targets is analyzed. Echo signal model is researched and derived, providing precondition for super-resolution algorithms.2. According to the Poisson distribution model, an improved Poisson maximum likelihood super-resolution algorithm is proposed. Introducing adaptive iteration acceleration factor, iterative convergence speed of the algorithm is effectively improved. Using prior information of amplitude statistics of targets, Poisson maximum a posteriori superresolution algorithm is proposed. It can obtain better super-resolution ability than the maximum likelihood algorithm.3. According to the Gauss distribution model, an improved Gauss maximum likelihood super-resolution algorithm is proposed. Using projection on convex sets method and introducing constraint conditions, the super-resolution ability is enhanced effectively. The maximum likelihood super-resolution algorithm based on mixed distribution is proposed. Using Poisson and Gauss mixed distribution model can suppress effectively of parasitic ripple and noise amplification, and improve its super-resolution ability.4. The super-resolution algorithm based on maximum entropy is proposed. It neglects amplitude distribution model, uses information entropy to measure the uncertainty of solution and chooses maximum entropy as regularize criterion. The algorithm can be applied in different types of distributions.5. According to the semiconvergence problem of super-resolution algorithms, the iterative stopping criterion based on residual signal is analyzed. Comparing mean square value of the residual signal with variance of noise, it timely terminates Gauss super-resolution algorithm. Stopping criterion based on random generalized cross validation is proposed. It compares value of random generalized cross validation of adjacent iterations, timely terminates the iteration process. The criterion has more extensively practical application.The above algorithms and criterions have been verified by simulations and real data, and can solve many problems of radar angular super-resolution.