Signal Processing Techniques for Landmine Detection Using Impulse Ground Penetrating Radar (ImGPR)

• Main Author: Tesfamariam, Gebremichael Teame
• Format: Others
• Language: English; en
• Published: 2013
• Online Access: https://tuprints.ulb.tu-darmstadt.de/3565/2/Gebremichael%20Te-ame%20thesis.pdf; Tesfamariam, Gebremichael Teame <http://tuprints.ulb.tu-darmstadt.de/view/person/Tesfamariam=3AGebremichael_Teame=3A=3A.html> (2013): Signal Processing Techniques for Landmine Detection Using Impulse Ground Penetrating Radar (ImGPR).Darmstadt, Technische Universität, [Ph.D. Thesis]

Landmines and unexploded ordinance (UXO) are laid during a conflict against enemy forces. However, they kill or maim civilians decades after the conflict has ended. There are more than 110 million landmines actively lodged in the globe. Every year more than 26,000 innocent civilians are killed or maimed. Most modern landmines are mainly nonmetallic or plastic, which are difficult to be detected using conventional metal detectors. Detection using hand-held prodding is a slow and expensive process. Impulse Ground Penetrating Radar (ImGPR) is a nondestructive technique capable of detecting shallowly buried nonmetallic anti-personnel (AP) and anti-tank (AT) landmines. In this PhD thesis, ImGPR is considered as a tool to detect landmines and UXO. The presence of strong ground clutter and noise degrade the performance of GPR. Hence, using a GPR sensor is almost impossible without the application of sophisticated signal processing. In electromagnetic wave propagation modeling, a multilayer transmission line technique is applied. It considers different soil types at different moisture levels. Plastic targets of different diameters are buried at different depths. The modeled signal is then used to estimate the ground and buried target parameters. In a parameter estimation procedure, a surface reflection parameter method (SRPM) is applied. Signal processing algorithms are implemented for clutter reduction and decision making purposes. Attention is mainly given to the development of techniques, that are applicable to real-time landmine detection. Advanced techniques are preceded by elementary preprocessing techniques, which are useful for signal correction and noise reduction. Background subtraction techniques based on multilayer modeling, spatial filtering and adaptive background subtraction are implemented. In addition to that, decorrelation and symmetry filtering techniques are also investigated. In the correlated decision fusion framework, local decisions are transmitted to the fusion center so as to compute a global decision. In this case, the concept of confidence information of local decisions is crucial to obtain acceptable detection results. The Bahadur-Lazarsfeld and Chow expansions are used to estimate the joint probability density function of the correlated decisions. Furthermore, a decision fusion based on fuzzy set is implemented. All proposed methods are evaluated using simulated as well as real GPR data measurements of many scenarios. The real data collection campaign took place at the Griesheim old airport and Botanischer Garten, Darmstadt, Germany in July 2011.