Study On Radar Target Recognition Algorithms Using High Resolution Range Profiles Based On Sparse Representation

The high resolution range profiles(HRRP)have the advantages of high resolution and contain physical structure information of the observed targets,which reflect the distribution of the target scattering centers along the radar line of sight.Compared with other multidimensional images,HRRP are much easier to be accessed and dealt with.They have been under the spotlight in radar automatic target recognition applications recently.But in some actual applications,radar systems are used to recognize those non-cooperative targets in complex environment.So we must deal with the problems,such as low SNR,small set of samples,feature extraction difficulty,complicate analysis,and so on.This research tries to combine sparse representation with RATR.In the dissertation,there are several main issues to be analyzed fully and deeply,which are the structure of RATR system,key techniques,experiments and tests.The main contents and the new ideas are as follows:1.By analyzing the formation mechanism and characteristics of HRRP and studying sparse representation theory,an innovation that uses sparse signal representation to recognize a radar target is proposed in the dissertation.The technical architecture and the key technologies are identified for this kind of recognition analysis.2.A recognition algorithm is proposed,which is based on the sparse representation of HRRP with a known structure dictionary.The recognition algorithm is discussed in three aspects.Firstly,the typical Gabor dictionary will be devided structurally into a number of atom subsets.Every atom subset can be represented by a key atom in it so to simplify the dictionary and all of the atom subsets will constitute a structual dictionary.Then,an improved fast matching pursuit algorithm is discussed.The IFMP algorithm is optimized with the globally optimal solution(GA)and fast operation to improve its convergence and the decomposition rate.At last,a ATR algorithm is proposed.The application model of sparse theory used in RATR is presented and is verified by simulation experiment.The experiments results show that the RATR algorithm in Chapter 3 is practicable,accurate and effective.In contrast to the traditional dimension reduction recognition method,the algorithm reduces computational complexity greatly and works well.3.A recognition algorithm is proposed,which is based on the sparse representation of HRRP with a federated dictionary.The recognition algorithm is studied in three aspects too.First,a federated dictionary is studied,which is constituted of several orthogonal sub-dictionaries.All of them have different expression characteristics.The whole dictionary is redundant and can be used to represent HRRP samples quickly and simply.Next,an improved Partition Matching Pursuit algorithm is discussed to help boost virtual sparse decomposition speeds.Then,a RATR algorithm is put forward on the basis of the first two steps.Moreover,some sparse decomposition parameter is adjusted by SNR in order to suppresses noise.The experiments results show that compared with the same kind of RATR algorithms,the algorithm in Chapter4 is practicable,simple and efficient.In contrast to the traditional dimension reduction recognition method,it has better noise robustness and higher recognition ratio.4.A recognition algorithm is proposed,which is based on the double sparse dictionary learning theory.An adaptive sparse representation method is used to analyze HRRP samples.The recognition algorithm is studied in two aspects.First,a double sparse analysis method is researched to dispose the dictionary and the HRRP samples respectively.Second,a stagewise OMP algorithm is used to improve the efficiency of the samples decomposition and representation.During the training phase,the target HRRP samples will be dynamically analyzed to get the category dictionary.During the training phase,the category dictionary will be used to test the other samples.Then,those samples can be identified according to the accuracy of their sparse representation results.The experiments reveal that this RATR algorithm is better than other similar algorithms.It can adaptively analyze those samples and acquire a more accurate category dictionary so it has better recognition performance.Compared with other typical RATR algorithms,the algorithm can ease the calculation strains on analyzing the samples for using the sparse analysis technique.The algorithm has the advantages of high measuring accuracy,better real time response,anti-noise,and is more practical.