Imaging And Discrimination Of BOR Targets Based On Compressive Sensing Theory

Based on the techniques of ultra wideband and synthetic aperture, the Ground Penetrating Radar(GPR) can effectively penetrate the ground surface to image and discriminate the shallow buried targets. Owing to the advantages of high efficiency and fast detection, the Forward Looking GPR(FLGRP) has been the research trend and emphases of GPR in the region of landmine detection. Compared with most targets, the landmine belongs to the special target with apriori knowledge. The discrimination approach based on the image is the matured procedure of FLGPR for landmine discrimination. Though this universal approach is suitable for landmine discrimination, the utilization of landmine’s apriori knowledge mainly concentrates on the feature extraction and discrimination, which is obviously inadequate. There seem to be lack of novel minds and approaches for landmine discrimination.This thesis is set in the landmine discrimination of FLGPR. Based on the analysis of landmine’s electromagnetic scattering characteristic, the brand new theory named compressive sensing(CS) is introduced. The research contents of this thesis gradually swerve form the conventional discrimination approach based on the image to the integrative approach of imaging and discrimination based on the sparse scattering model. Accordingly, the research object swerves from the landmine to the normal Body-of-Revolution(BoR) target. All the researches in this thesis are bran-new attempts of CS in discriminating the target with apriori knowledge.Firstly, based on the techniques of Stepped Frequency(SF) and Split Aperture Transmit Configuration(SATC) the imaging model of the FLGPR system is introduced, and then the Back Projection(BP) algorithm which is considered as the kernel of FLGPR’s 2-dimensional imaging algorithm is discussed. Due to the ultra bandwidth and multiple channels, it is difficult to use single calibrator to compensate the system errors. For solving this problem, a new system compensation approach based on the confusion of multiple calibrators is proposed, which effectively calibrates the system errors. Moreover, in order to solve problem of image defocusing induced by the sensitivity of SF in movement and the module reuse of SATC, a new approach of speed estimation and compensation is proposed, which is based on the optimal estimation of the contrast ratio of the Region Of Interest(ROI) image. Both the above approaches effectively improve the image quality of the FLGPR system.Secondly, the approach of CS to robustly reconstruct the noisy signal is well researched. Aiming at achieving the robust reconstruction of CS with Single Random Measurement(SRM) in low Signal-to-Noise Ratio(SNR) conditions, a novel robust reconstruction approach based on multiple measurement iterative pixel discrimination is proposed. This new approach not only lays the foundations for the practical application of CS theory, but also becomes a key support to the sparse imaging and discrimination of CS in this thesis.Thirdly, based on the CS theory, both the extraction and discrimination of landmine’s sparse scattering structure are investigated. With the approach of electromagnetic calculation, the landmine’s electromagnetic scattering characteristic(ESC) is analyzed. According to the landmine’s azimuth-invariable double-hump structure, a new approach based on the CS theory is proposed for extracting the sparse scattering structure of landmine. Via CS processing, the utilization of landmine’s apriori knowledge is brought forward to the imaging, and the landmine discrimination is changed into the discrimination of geometrical structure, which comes to an integrative processing of sparse imaging and target detection. According to the above conclusions, a novel approach for landmine discrimination is proposed. Based on the sparse image of the minefield, the landmine can be discriminated by the geometrical features extracted from landmine’s apriori knowledge.Fourthly, the integrative approach of sparse imaging and target discrimination is investigated. With the Feature-based Dictionary(FD) constructed with the landmine’s apriori Discrete Scattering Structure(DSS), both the sparse imaging and discrimination of the landmine can be received simultaneous by CS processing. Moreover, the FD-based integrative approach is applied to the BoR targets with DSS. In order to overcome the limitation of FD on discriminating the BoR targets with similar or no DSS, a novel integrative approach which is based on the Model-based Dictionary(MD) constructed with the target’s ESC is proposed. Compared with the FD-based approach, the MD-based approach can effectively discriminate the BoR target, no matter the target has DSS or not. Consequently, the MD-based discrimination approach not only breaks through the conventional discrimination procedure, but also attempts to a new way for discriminating the targets with apriori knowledge.Finally, both the CS-based radar and its applications are discussed. Based on the techniques of SF and SATC, a CS-based radar system with Multiple-Input and Multiple Output(MIMO) is designed. In order to demonstrate the feasibility and applicability of the proposed SF-MIMO-CS radar system, it is applied to the Down Looking GPR(DLGPR) and Through-Wall Imaging(TWI). With the foregoing research results in this thesis, both the DLGPR and TWI receive the satisfying results.