Pambudi, Afief Dias (2021)
Robust Signal Processing for Landmine Detection with Forward-Looking Radar.
Technische Universität Darmstadt
doi: 10.26083/tuprints-00019672
Dissertation, Erstveröffentlichung, Verlagsversion
Kurzbeschreibung (Abstract)
Landmines and other unexploded ordnance are one of the greatest curses in modern times, the legacies of war continuing to pose threats to civilians many years after a conflict has ended. They limit freedom of movement and deny access to basic human needs, which hinder post-conflict reconstruction efforts and the implementation of the sustainable development goals of a country. Detection and clearance of landmines is a very risky operation, which requires high standards on technology and methodology.
This PhD thesis is dedicated to the problem of landmine detection using forward-looking ground penetrating radar in rough surface environments. Robust signal processing techniques are developed, which offer a guaranteed performance of landmine detection with the forward-looking radar. A multi-view imaging approach is used in which tomographic radar images are obtained from multiple viewpoints of the investigation area.
An approach based on a simple threshold and likelihood-ratio test is first examined for multi-view image fusion and detection. The imaging of a complex scene containing nine plastic and metallic targets is considered. The detection problem is defined as a test between a null (mines absent) and an alternative (mines present) hypothesis. A parametric family of distribution models is employed to obtain nominal distributions of the pixel intensity of the radar image under both hypotheses. The nominal distribution models are then used for fusion schemes based on the likelihood-ratio test detector.
In a robust approach, uncertainty distribution models under both hypotheses are considered. The detector is designed to be robust against the deviations of distributions. Two density bands are constructed within which the corresponding probability density functions of feasible distributions are assumed to lie. The detector is then designed such that it minimizes the maximum error probability for all possible density pairs within the two bands. The reason behind following a robust minimax approach is that accurate estimation of the distribution, given the varying nature of the interrogated environment, is highly challenging. The robust techniques overcome this issue since they do not need an accurate initial estimate and are guaranteed to perform well over a set of feasible distributions.
The design of the detector is further extended by accounting for the statistical dependency between multiple radar images from different viewpoints of the investigation area. Different copula density functions are investigated in terms of their effectiveness in incorporating dependence between multi-view images into the test statistics. It is shown that incorporating statistical dependence with a well-selected copula model improves the test performance of both non-robust and robust techniques, which affirms the benefit of the implementation.
Typ des Eintrags: | Dissertation | ||||
---|---|---|---|---|---|
Erschienen: | 2021 | ||||
Autor(en): | Pambudi, Afief Dias | ||||
Art des Eintrags: | Erstveröffentlichung | ||||
Titel: | Robust Signal Processing for Landmine Detection with Forward-Looking Radar | ||||
Sprache: | Englisch | ||||
Referenten: | Zoubir, Prof. Dr. Abdelhak M. ; Ahmad, Asc. Prof. Fauzia | ||||
Publikationsjahr: | 2021 | ||||
Ort: | Darmstadt | ||||
Kollation: | xi, 97 Seiten | ||||
Datum der mündlichen Prüfung: | 15 September 2021 | ||||
DOI: | 10.26083/tuprints-00019672 | ||||
URL / URN: | https://tuprints.ulb.tu-darmstadt.de/19672 | ||||
Kurzbeschreibung (Abstract): | Landmines and other unexploded ordnance are one of the greatest curses in modern times, the legacies of war continuing to pose threats to civilians many years after a conflict has ended. They limit freedom of movement and deny access to basic human needs, which hinder post-conflict reconstruction efforts and the implementation of the sustainable development goals of a country. Detection and clearance of landmines is a very risky operation, which requires high standards on technology and methodology. This PhD thesis is dedicated to the problem of landmine detection using forward-looking ground penetrating radar in rough surface environments. Robust signal processing techniques are developed, which offer a guaranteed performance of landmine detection with the forward-looking radar. A multi-view imaging approach is used in which tomographic radar images are obtained from multiple viewpoints of the investigation area. An approach based on a simple threshold and likelihood-ratio test is first examined for multi-view image fusion and detection. The imaging of a complex scene containing nine plastic and metallic targets is considered. The detection problem is defined as a test between a null (mines absent) and an alternative (mines present) hypothesis. A parametric family of distribution models is employed to obtain nominal distributions of the pixel intensity of the radar image under both hypotheses. The nominal distribution models are then used for fusion schemes based on the likelihood-ratio test detector. In a robust approach, uncertainty distribution models under both hypotheses are considered. The detector is designed to be robust against the deviations of distributions. Two density bands are constructed within which the corresponding probability density functions of feasible distributions are assumed to lie. The detector is then designed such that it minimizes the maximum error probability for all possible density pairs within the two bands. The reason behind following a robust minimax approach is that accurate estimation of the distribution, given the varying nature of the interrogated environment, is highly challenging. The robust techniques overcome this issue since they do not need an accurate initial estimate and are guaranteed to perform well over a set of feasible distributions. The design of the detector is further extended by accounting for the statistical dependency between multiple radar images from different viewpoints of the investigation area. Different copula density functions are investigated in terms of their effectiveness in incorporating dependence between multi-view images into the test statistics. It is shown that incorporating statistical dependence with a well-selected copula model improves the test performance of both non-robust and robust techniques, which affirms the benefit of the implementation. |
||||
Alternatives oder übersetztes Abstract: |
|
||||
Status: | Verlagsversion | ||||
URN: | urn:nbn:de:tuda-tuprints-196722 | ||||
Sachgruppe der Dewey Dezimalklassifikatin (DDC): | 600 Technik, Medizin, angewandte Wissenschaften > 600 Technik 600 Technik, Medizin, angewandte Wissenschaften > 620 Ingenieurwissenschaften und Maschinenbau |
||||
Fachbereich(e)/-gebiet(e): | 18 Fachbereich Elektrotechnik und Informationstechnik 18 Fachbereich Elektrotechnik und Informationstechnik > Institut für Nachrichtentechnik 18 Fachbereich Elektrotechnik und Informationstechnik > Institut für Nachrichtentechnik > Signalverarbeitung |
||||
Hinterlegungsdatum: | 14 Okt 2021 12:20 | ||||
Letzte Änderung: | 15 Okt 2021 07:02 | ||||
PPN: | |||||
Referenten: | Zoubir, Prof. Dr. Abdelhak M. ; Ahmad, Asc. Prof. Fauzia | ||||
Datum der mündlichen Prüfung / Verteidigung / mdl. Prüfung: | 15 September 2021 | ||||
Export: | |||||
Suche nach Titel in: | TUfind oder in Google |
Frage zum Eintrag |
Optionen (nur für Redakteure)
Redaktionelle Details anzeigen |