Font Size: a A A

Study On Electro-magnetic Feature Extraction Of SAR/ISAR And Its Applications

Posted on:2016-09-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:J DuanFull Text:PDF
GTID:1108330482453145Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
Radar is a means of remote sensing that uses electro-magnetic waves to obtain the electro-magnetic features of targets. Different from other remote sensing means such as traditional optics, radar has the ability to observe in all types of weather and day/night, survey at long ranges and penetrate optically opaque surfaces. However, it is diffcult for people who are familiar with visual light data and photography to understand electro-magenetic radar images. Moreover, even the analyzers of radar images may come to different conclusions based on images of same targets. Therefore, we propose an adaptive radar interpretation method to understand radar echoes and images. As bridge between radar data and useful information, it makes sense to extract electro-magentic features of radar targets to interpret radar images. Gernerally speaking, the electro-magentic features of radar is consisted of two parts:the features of scattering centers and the polarimetric features. Based on electro-magentic mechanisms, parametric models can be built to describe the behaviors of scatteirng centers with parameters corresponding to specific physical meanings. In this way, feature extraction of scattering centers can be completed by estimating the parameters of parametric models from radar echoes. Moreover, modern radar is equiped with multi-frequency and multi-polarization. The polarimetric radar is capable of obtaining throrogouh information of targets by transmitting and receiving echoes with two orthogonally polarized antennas. The polarimetric features can be obtained by applying polarimetric target decompostion (PTD) to the scattering matrix of pixels on basic scattering mechanisms. The polarimetric features are related to structures and materials, and thus can be used through inversion to measure surface parameters and classify targets, etc.This dissertation studies electro-magnetic feature extraction of manmade targets to improve comprehension of radar images and the constrast of targets from noise by means of signal processing. The three main points are:feature extraction, applications of features such as target detection, classification and super-resolution etc, and the extended application to transient interference excision of the Over the Horizon radar (OTHR). The relevant work is supported by the National Basic Research Program of China (973 Program), National Science Foundation of China under grant No.61101245 and No.61301280, and the Fundamental Research Funds for the Central Universities under grant K5051302001,K5051302078 and K5051302038.The main content of this dissertation is summarized as follows.The first part focuses on feature extraction of targets based on electro-magnetic mechanisms. First, the typical scattering center models based on electro-magnetic mechanisms are built, whose relationships are analyzed as well. By analyzing the effects of parameters, a parameter decoupling and estimation method for independent scattering centers is proposed to cut down the intensive computation caused by high dimensional parameters. By means of sample selection, the estimation in a Gaussian noise environment can be approached as the estimation in ideal cases without noise. Afterwards, a parameter estimation method is proposed for radar images of multiple scattering centers based on the RELAX algorithm. This way, the intensive computation of estimation can be reduced. Problems in determining total scattering center numbers can be avoided, as well. Moreover, the minimum searching intervals of parameters are deduced. In order to minimizes the iterations to converge using the RELAX based parameter estimation method, the orthogonal matching pursuit method is brought in to update the parameters for each iteration by renewing the dictionary. However, since a continuous structure may behave discontinuous in radar image under different polarizations, estimation errors always occur. Therefore, a parameter estimation method with joint polarized data is proposed by making use of the same features across different polarizations.In the second part, we study the applications of extracted electro-magnetic features. First, a target detection method is presented to detect manmade targets based on the extracted features of scattering centers. Different from the traditional pixel based detection, this method can preserve the integrity of structures and robustness to noise by considering the correlation of pixels belonging to same structures. Finally, by making use of the extracted polarimetric parameters, the scattering matrix of scattering centers can be constructed. After applying the polarimetric target decomposition to the scattering matrix, the scattering mechanisms of each scattering centers can be analyzed. Since the matrix of scattering center is analyzed instead of that of pixel in this proposal, it can be utilized to perform polarimetric target decomposition to manmade targets, which fails traditional methods with structure discontinuity and sensitivity to noise.In the final part, an extended application of scattering center models based on electro-magnetic scattering mechanisms is studied. Obviously, the signal model of targets can be constructed by the extracted parameters, through which the stored data have been compressed greatly. Moreover, we can extrapolate the observing frequency and aspect angle by making use of the reconstructed signal model, resulting in super-resolution radar images. Compared with available super-resolution methods, both the continuity of structures and robustness to noise can be improved by the method proposed in this paper. Since the transient interference of OTHR is of short time duration and intensive amplitude in time domain, it behaves as a sine-like function in the Doppler domain. Hence, a rectangular window base (RWB) has been proposed to describe the physical behaviors of transient interference in the time domain, which is similar to the attributed scattering centers in the Doppler domain. Based on sparse decomposition, the OTHR radar has been decomposed as sums of RWBs. By making use of the fast Fourier transform, this operation can be fast. After applying thresholds to the extracted features, dual-threshold CFAR detection is completed to separate the transient interference from the signal, which is independent of the coherent integrate time of radar. Moreover, a Doppler spectrum reconstruction method of target and clutter based on compressive sensing is described for sparse sampling cases, in which targets are submerged by high grating lobes.
Keywords/Search Tags:Electro-magnetic Feature Extraction, Target Detection, Super-resolution, Polarimetric Target Decomposition, Transient interference Excision
PDF Full Text Request
Related items