Advances in GPR-based landmine automatic detection

https://doi.org/10.1016/j.jfranklin.2009.02.014Get rights and content

Abstract

As an application of mechatronics, this paper presents the advances in surface-adaptive ground penetrating radar (GPR)-based anti-personnel landmine detection project in Nagoya University. These advances can be summarized in three items: (1) GPR manipulation where a low-pressure-tire vehicle capable of moving inside a mine field, to facilitate machine-based sensing in place of manual sensing, is applied; (2) enhancement of underground landmine suspects’ images through geography adaptive scanning and measurements signal processing of a vector frequency modulated continuous wave (FMCW) GPR; (3) GPR fusion with metal detector (MD) for automatic decision making through experimental-based fuzzy learnt fusion rules. The state-of-art of these advances as well as directions for future research work is to be presented.

Introduction

According to the United Nations, as of the year 2000 there were 70 million landmines planted in a third of the world's nations affecting global causality rate of up to 20,000/yr [1]. That is why landmine detection has attracted much attention by many research teams around the world during the last few years; among them is our research team in Nagoya University. anti-personnel (AP) mine ranges from 5 to 15 cm in size; they can be metal, plastic, or wood. AP mines are normally buried at shallow depth; detonated by very low pressure, and designated to kill or maim people. PMN2, M14 and Type72 are examples. In real world clearance activities, AP mine suspect areas are divided into 1 m grid squares, and each square meter is probed with a bayonet or plastic rod. Probing is done at an oblique angle to the ground so that the rod will encounter the side of a land mine and not trip the fuse. No need to say, this work is very dangerous and proceeds very slow [2]. The need for a safer and more fast humanitarian demining action by replacing a manual sensing task by vehicle sensing task have motivated our research team to introduce a low-ground-pressure tires detection vehicle [3]. The unmanned vehicle can move in mine field without detonating a group of AP mines.

One of the big challenges in demining process is detection. If a mine is detected, deminers can explode, mark or move it to a pit for later detonation or defusing. Conventional mine detection, by a metal detector, is often difficult for two reasons. First, mines are increasingly being made of plastics, minimizing the more easily detectable metal components. Second, mined areas are often equipped with metal scraps creating a high false alarm rate. Because of the difficulty encountered in detecting the tiny amounts of metal in a plastic landmine with a metal detector, technology development has been extended to other sensors. Ground penetrating radar (GPR) used for about 70 years for a variety of geophysical subsurface imaging applications including utility mapping and hazardous waste container has been actively applied to the problem of land mine detection for nearly the last two decades of research. It provides sensing objects underground based on dielectric properties. It senses the reflected electromagnetic wave by a buried object. It is expected that GPR be a good alternative sensor and/or an important support sensor when fused with a metal detector for landmine detection. However, one major source of error in GPR data is the reflection from the surface of the ground [4]. The problem becomes much more difficult for an undulating ground surface. As a general objective of signal processing as applied to GPR is to present an image that can be easily interpreted by the operator, it is important to adapt the signal processing technique for an undulating surface scanning. In this paper, signal processing for ground-surface-adaptive scanning [5], applying a vector GPR [6], will be presented.

A metal detector is one of the most major sensors applied for current humanitarian demining. It is simple and cost effective. It is also reliable to find an anti-personal mine (APM) in a shallow subsurface. However, it suffers from the high false alarm rate (about 99.95%), as it senses all metal objects including metal fragments in the field other than land mines. On the other hand, ground penetrating radar provides (after processing), images for objects underground based on dielectric properties. However, it senses a land mine object as well as any other object as it senses dielectric discontinuities in metallic and/or non-metallic objects. Fusion of GPR with MD is expected to minimize the false alarm rate significantly. In this research, fusion of both MD and GPR for APM detection in a shallow subsurface is presented. A “feature in-decision out” fuzzy sensor fusion algorithm for GPR and MD is introduced [7]. The inputs to the fuzzy fusion system are features extracted from both GPR and MD measurements. The output from the fuzzy fusion system is a decision if there is a land mine and at what depth it would be. Fuzzy fusion rules are extracted from training data through a fuzzy learning algorithm. Experimental test results are presented to demonstrate the validity of the proposed fuzzy fusion algorithm and hence its influence in minimizing the false alarm rate for mine detection.

This paper is organized as follows: Section 2 introduces GPR manipulation where a low-pressure-tire vehicle capable of moving inside a mine filed is presented. Section 3 presents the enhancement of landmine images through signal processing of a scanned undulating surface applying a vector FMCW GPR. Section 4 presents the fuzzy fusion algorithm of GPR with MD as well as experimental evaluation of the learnt fusion rule base. Section 5 presents conclusions and projected work.

Section snippets

GPR manipulation

In this project, we propose the landmine detection system which has the following features based on the problems encountered in previous demining systems: (1) can enter mine field; (2) uses compound sensor; (3) sensing with adaptation to landform and vegetation; (4) records sensing result to information management system and uses it for prodding or other mine action. The prototype system consists of an access vehicle which enters a mine field and an operator's instrument used for handling

GPR images enhancement for geography adaptive scanning

It is known that frequency modulated continuous wave FMCW radar systems have been used in preference to AM systems where the targets of interest are shallow and frequencies above 1 MHz can be used [4]. Since a GPR response signal is reflection intensity versus time, image reconstruction is required for easier extraction of mine suspects, for both easier interpretation and automatic detection [8]. For this purpose, many signal processing methods have been proposed and applied to GPR imaging, but

GPR–MD fusion

In this section, a “feature in-decision out” fuzzy sensor fusion algorithm for GPR, and a metal detector (MD), for mine detection is introduced. The inputs to the fuzzy fusion system are features extracted from both GPR and MD measurements. The output from the fuzzy fusion system is a decision if there is a land mine and at what depth it would be. Fuzzy fusion rules are extracted from training data through a fuzzy learning algorithm. Experimental test results are presented to demonstrate the

Conclusions

The advances in surface-adaptive ground penetrating radar (GPR)-based anti-personal landmine detection project in Nagoya University is presented. GPR manipulation by a low-pressure-tire vehicle capable of moving inside a mine field without detonating a group of anti-personnel landmines is presented. Signal processing for geography adaptive scanning applying a vector GPR is presented. Enhancement of underground landmine suspects’ images is experimentally verified. GPR fusion with metal detector

References (15)

  • S.L. Anderson

    Landmine detection research pushes forward despite challenges

    IEEE Intelligent Systems Magazine

    (2002)
  • R. Siegel

    Land mine detection

    IEEE Instrumentation & Measurement Magazine

    (2002)
  • Y. Hasegawa, Y. Kawai, K. Yokoe, T. Fukuda, Low-ground-pressure vehicle for adaptive mine detection, in: Proceedings of...
  • D.J. Daniels

    Ground Penetrating Radar

    (2004)
  • Y. Hasegawa, K. Yokoe, Y. Kawai, T. Fukuda, GPR-based adaptive sensing, in: Proceedings of the 2004 IEEE/RSJ...
  • T. Fukuda, Y. Hasegawa, Y. Kawai, S. Sato, Z. Zyada, T. Matsuno, Automatic land-mine detection system using adaptive...
  • Z. Zyada, Y. Kawai, T. Matsuno, T. Fukuda, Fuzzy sensor fusion for mine detection, in: Joint 3rd International...
There are more references available in the full text version of this article.

Cited by (14)

  • Performance assessment of SVM-based classification techniques for the detection of artificial debondings within pavement structures from stepped-frequency A-scan radar data

    2019, NDT and E International
    Citation Excerpt :

    reviews a three-decade journey of the GPR application in Civil engineering from testing and evaluation to diagnosing the pavement structures. In the field of pavement monitoring, GPR has been used to estimate the thickness of pavements, detect debondings [10–12] and identify buried objects such as anti-personnel mines [13,14], steel bars in concrete [15,16] etc. However, as mentioned in [3], each of these applications require suitable processing methods to help at data interpretation and decision making (classification, identification etc.).

  • Symplectic partitioned Runge-Kutta methods for two-dimensional numerical model of ground penetrating radar

    2012, Computers and Geosciences
    Citation Excerpt :

    Ground penetrating radar (GPR), as a kind of high resolution, non-destructive tool, has been widely applied to various disciplines, such as archaeology (Böniger and Tronicke, 2010), civil engineering (Xu et al., 2010), and mine detection (Zyada et al., 2011).

  • GPR-signal improvement using a Synthetic Emitter Array

    2011, Journal of Applied Geophysics
    Citation Excerpt :

    The acquisition of dense grids of parallel lines can provide detailed maps of targets (Francese et al., 2009; Hugenschmidt and Kalogeropoulos, 2009; Pérez-Gracia et al., 2009). SO methodologies have been applied to archeological prospection, pipe and tank detection, land-mine detection, engineering-structure characterization, forensic investigation, groundwater and leakage monitoring and shallow-geomorphology mapping, etc. (Bhosle et al., 2007; Bonomo et al., 2009; Fiedler et al., 2009; Gomez et al., 2009; Orlando and Slob, 2009; Pettinelli et al., 2008; Porsani and Sauck, 2007; Zyada et al., 2011). In this paper we discuss the use of a synthetic antennae array to improve GPR signals.

  • Pipe-flange detection with GPR

    2011, Journal of Geophysics and Engineering
View all citing articles on Scopus
View full text