Application of Time-Frequency Transformations in Polarimetric Ultra-Wideband MIMO-GPR signals for Detection of Colombian Improvised Explosive Devices

Abstract. In this thesis, a new radar technique for GPR detection and discrimination of Improvised Explosive Devices is presented and validated. Data processing, consisting of adaptive filters and time-frequency transformations, are applied to polarimetric GPR data, in order to construct feature vectors of the targets. These vectors are used as inputs of a support vector machine algorithm, in order to discriminate buried targets either as improvised explosive device (IED) or clutter. The main contributions of this thesis are as follows. First, the permittivity of improvised ANFO explosives is measured. This information is used for manufacturing inert surrogates of IEDs. Second, we proposed the construction of target feature vectors (TFVs) from polarimetric GPR measurements. Third, recursive algorithms and background removal are combined to improve the clutter removal. Data processing methods are assembled, combining clutter removal stage, time-frequency transformation and singular value decomposition. In total, eight data processing methods are proposed. Moreover, for every method, 13 TFVs are assembled. Then, the TFVs are used to train and test support vector machines (SVM) under a binary classification approach. Classification results are validated by using the leave-two-out cross-validation. Accuracy of 87.02% in the best classifier was obtained. The main conclusion of this thesis is that combining polarimetric GPR measurements, feature extraction using time-frequency transformations, and SVM classifications allows obtaining discriminating features that improve the IED detection rates compared with metal detector performance. Furthermore, the proposed approach can be implemented in a hand-held detection device and to be used in a humanitarian demining scenario. Keywords: Classification of improvised explosive devices, feature extraction, ground penetrating radar, permittivity of explosives, polarimetric measurements, support vector machines, ultra-wideband MIMO radar.

• Autores: Gutiérrez Duarte, Sergio Alonso
• Tipo de recurso: Doctoral thesis
• Fecha de publicación: 2019
• Institución: Universidad Nacional de Colombia
• Repositorio: Universidad Nacional de Colombia
• Idioma: spa
• OAI Identifier: oai:repositorio.unal.edu.co:unal/76613
• Acceso en línea: https://repositorio.unal.edu.co/handle/unal/76613; http://bdigital.unal.edu.co/73192/
• Palabra clave: Classification of improvised explosive devices; Landmine detection; Ground penetrating radar; Feature extraction; Polarimetric measurements; Support vector machines; Ultra-wideband MIMO radar; Polarimetric radar; Machine learning
• Rights: openAccess
• License: Atribución-NoComercial 4.0 Internacional