Radar Target Recognition And Superresolution Imaging Research

Automatic Target Recognition(ATR)based on radar imaging plays an important role in modern engineering applications and has become a very hot research in the world.For target recognition,the target has some specific information and some feature vectors can be extracted.In accordance with the radar resolution performance,ATR can be divided into low resolution and high resolution target recognition.Low resolution echo signal for target recognition is limited which can only make rough estimation.However,high resolution image contains more characteristic information of the target,ways to improve image resolution and realize efficient target extraction are extensively investigated in the field of identification.The general high resolution target recognitions include High Resolution Range Profile(HRRP)and Synthesized Aperture Radar(SAR)image or inverse synthetic aperture radar(ISAR)image.In recent years,the technology of HRRP,SAR and ISAR has been mature,which provides strong support for the development of automatic target recognition technology.In this dissertation,we investigate target recognition based on high resolution range profile from radar,SAR and ISAR.Firstly,we research the target recognition based on ISAR HRRP and MSTAR SAR image;Then using ultra wideband(UWB)signal to improve range resolution is very effective.With respect to the previous work,this dissertation focuses on the following aspects:1.Firstly,based on analysis of the imaging principle and establishment of SAR and ISAR model,range cell migration correction(RCMC)for SAR imaging and envelope and phase error caused by translational motion for ISAR imaging are done with the help of some classical techniques,such as,minimum entropy method(MEM)and Phase Gradient Autofocus(PGA).Method of equivalent edge currents(MEC)is used to calculate the scattering of aircrafts.With adoption of the Range Doppler(RD)algorithm,favorable images of the aircrafts are obtained.2.In the field of target recognition,a new noise-robust method for the radar HRRP reconstruction via unitary ESPRIT algorithm is proposed.The main idea of the unitary ESPRIT algorithm is to transform the complex-valued data to real-valued data,which can improve the data utilization and reduce the computational complexity.Comparing with the conventional ESPRIT algorithm,the unitary ESPRIT algorithm has higher resolution and better anti-noise ability.Then reconstruct the HRRP by means of the extracted scattering centers,i.e.locations and amplitudes.Experimental results show that the proposed classification scheme has better robustness,recognition rate and anti-noise than conventional method which base on IFFT.3.A new method employing active contour without edges is presented for SAR target recognition.Since the SAR image contains lots of background noise which will affect the accuracy of radar target recognition,the active contour without edges extraction technique from the perspective of image model gives the global energy functional that the image model should satisfy.This method has many advantages such as strong anti-noise ability and fast iterative convergence speed.Then we use Hu invariant moments to feature extraction and Support Vector Machine(SVM)for classification.The experimental data used for classification was MSTAR database that was collected using the Sandia National Laboratories Twin Otter SAR sensor payload operating at X band.Moreover,from the comparison results with conventional method,this method is more stable and faster and has higher recognition rate under different training samples.4.In order to obtain HRRP,an ultra wideband signal method by coherently combining multi-bands measurement is presented.First,the all-pole model for different subbands are estimated respectively.Then,we estimate the pole value of subbands using the root-MUSIC algorithm and the linear least-square method is used to estimate the pole model coefficients of the subbands.Non-linear least-square is employed to obtain global all-pole signal model.Finally,the effectiveness of the proposed method is applied into SAR imaging to verify its resolution improvement by means of simulated targets.5.The most effective way to improve the accuracy of target recognition is to increase the resolution of the image to obtain more target information.In order to improve the resolution,ultra wideband(UWB)LFM signal is transmitted.Meanwhile,to obtain the physical size of the target in azimuth direction,the parameters of the echo signal are estimated by some methods.To solve this problem,a parameter estimation algorithm for UWB LFM signal based on compressed sensing(CS)is discussed.This method uses less data and reduce running time significantly,which is good for target recognition.