de
en
Schliessen
Detailsuche
Bibliotheken
Projekt
Impressum
Datenschutz
zum Inhalt
Detailsuche
Schnellsuche:
OK
Ergebnisliste
Titel
Titel
Inhalt
Inhalt
Seite
Seite
Im Dokument suchen
Giovanneschi, Fabio: Online dictionary learning for classification of antipersonnel landmines using ground penetrating radar. 2020
Inhalt
Acknowledgments
Zusammenfassung
Abstract
Contents
Introduction
GPR and the de-mining problem
The Compressive Sensing framework
Major Contributions
Thesis outline
Ground Penetrating Radar Technology and Target Classification
Geophysical Background
GPR principles
GPR target classification
Overview and challenges
Analysis of GPR scattering signatures from landmine-like targets
Time domain analysis
Frequency domain analysis
Time-Frequency domain analysis
Study of the scattering responses
Sparse Representation and Dictionary Learning techniques
Fundamentals of Sparse Representation
Null space condition, coherence and Restricted Isometric Property (RIP)
Sparse Representation methods
Greedy approaches
Orthogonal matching pursuit (OMP)
A faster implementation of OMP
Batch-OMP
Convex optimization approaches
Fundamentals of Dictionary Learning
Dictionary Learning methods
Batch DL approaches
K-SVD
LRSDL
Online DL approaches
ODL
CBWLSU
DOMINODL
Online dictionary learning for adaptive GPR target recognition
Experimental measurements
Ground Penetrating Radar System
Test Field
Dataset Organization
A note on signal pre-processing
Selecting the maximal residual error for SR
Parametric evaluation of DL algorithms
Similarity Measure
Statistical Metrics
Coefficient of variation
Two-sample Kolmogorov-Smirnov (K-S) distance
Dvoretzky-Kiefer-Wolfowitz (DKW) inequality
Parametric Evaluation results
Influence of the number of iterations
Influence of the number of trained atoms
DOMINODL input parameters selection
Considerations on computational efficiency of DL algorithms
Classification results
Classification with Optimal Parameters
Classification with Non-Optimal Parameters
Comparison with Deep Learning Classification
Classification with Reduced Range Samples
Summary and conclusions
Existing challenges and proposed methodology
Evaluation of DL algorithms
Classification performances
Final remarks and outlook
Bibliography
Convolutional Neural Networks
Support Vector Machines
List of Figures
List of Tables