Novel Imaging Radar Technology for Detection of Landmines and Other Unexploded Ordnance
The safe, reliable, and efficient detection and subsequent removal of buried landmines or other unexploded ordnance (UXO) still remains a challenge. According to Landmine Monitor (2015), almost 60 states suffer the threat of antipersonnel mine contamination, while the total extent of contaminated areas is likely to be thousands of square kilometers and the total number of threat objects may be well over 100 million. In contrast, total annual clearing rates are in the order of about 230,000 landmines and about 200 km2 for the year of 2014, as the number of new contaminations continues to grow. Although non-technical surveillance for the identification of contamination improves slowly, more progress is needed in the basic detection of UXO. Although many new technologies have been investigated the last 20 years, the classical ones like metal detectors and dogs in cooperation with humans operators are still the most commonly ones used today. Hence, the detection process for many scenarios is unacceptably slow and dangerous for the operators, as they are typically less than 1 m away from the threats. Hence, a sensor technology enabling a sufficient stand-off distance for operators and reliable detection at high area throughput is desirable. The following paper describes the physical background, the system design, and representative measurement results of our innovative radar approach to this problem.
KeywordsSecurity research Landmine detection Unexploded ordnance Synthetic aperture radar Spatial resolution Radar clutter
Here the sensor is moving parallel to and along the contaminated area in sufficiently safe distance on safe ground. Using, for instance, a ground vehicle, the sensor is elevated above the scene to several meters and the stand-off range should be in the order of several meters, providing sufficient slant visibility on the contaminated area. The width of the actually observed contaminated area should be in the order of several meters, and the length is defined by the selected path length of the moving sensor. Such sensors are optimized for maximum and reliable detection performance of all objects in the scene, having not to discriminate in situ between true threats and false targets. Once a preferably geo-referenced detection map has been created, only the locations of detections need to be investigated further by applying a local suite of specific metal, chemical, GPR sensors in close proximity. The close-in secondary sensors can also be mounted on a steerable ground-based vehicle and optimized for detection and confirmation.
However, it is evident that a single system is not optimum for every scenario. The differing scenarios of a dense forest or jungle, rice field areas, densely covered urban structures, or strongly ragged terrain, will require specifically tuned sensor configurations. Nevertheless, there are many contaminated areas where the proposed approach can deliver considerable added value. A very suitable sensor type for that task is synthetic aperture radar (SAR). SAR techniques for landmine detection have been refined over a number of years (Lawrence et al. 1999; Liu et al. 2003; Lloyd and Longstaff 2003; Feng et al. 2005; Wu et al. 2010; Kabourek et al. 2012).
2 Basics of Imaging Microwave Radar
The interaction of electromagnetic waves with matter is dictated by the electric and magnetic properties of the material. The bulk is defined in terms of reflection, transmission, and absorption of incident energy. Measuring such parameters by sensors can deliver information about an object. A suitable frequency range to perform such observations with respect to sensor size, interaction responsivity, and hardware availability is the microwave region. Here the interaction is purely defined on the bulk behavior of the material, i.e., that determined by the permittivity, and the permeability of the material. Spectral line features generally appear in the terahertz, infrared, and optical region of the electromagnetic spectrum, but radiation in those bands has insufficient penetration for buried UXO detection. Microwaves, on the other hand, with their longer wavelengths have sufficient penetration in and through many non-metallic materials, so buried objects with some reflectivity will generate a signature. Hence in theory, even objects buried in natural soils can interact with incident waves, thus providing information on them. The microwave frequency range is roughly defined by 1–30 GHz, and the range of 30–300 GHz is called the millimeter-wave range. Corresponding wavelengths are 30–1 cm and 10–1 mm. Detailed information on radar and electromagnetic waves can be found in Ulaby (1981), (1982).
2.1 Reflection and Scattering
Consequently, by scanning a surface as indicated in Fig. 3, surface areas or even buried target areas can be detected by microwave radar purely based on mismatches in wave impedance, i.e., differences in permittivity and permeability between interfaces. However, those articulate weaker or stronger, heavily dependent also on previously mentioned other parameters, so that a few very important constraints on radar imaging have to be considered.
2.2 Penetration Depth
2.3 Spatial Resolution
It becomes evident that the spatial resolution of the imaging radar should achieve at least the size of the smallest objects to be detected. Otherwise, as shown in the low resolution image, targets may not be discriminated spatially from each other and the contrast with respect to the surrounding background decreases considerably. Hence a successful detection may fail. A reasonable requirement for the spatial resolution to be achieved by the radar was estimated to be in the order of 5–10 cm.
2.4 Backscatter of Targets and Background Clutter
In the optical image, a green ball is located on a green grass area. Although the ball can be made of completely different material like plastics, the color of the ball and the grass here is rather similar, so that a proper discrimination of the ball from the grass background is hard. A similar situation can occur to radar imaging when a target is located in a cluttered background. Figure 6 right shows average normalized RCS values of different terrain in dependence of incidence angle for two useful frequencies in landmine detection. Assuming now landmines or UXO as spherical and mostly dielectric objects, their RCS levels can be approximated by metallic or dielectric spheres of similar size. For those objects, the RCS can be computed and normalized values around 10−3–10−4 or even lower can be obtained at those frequencies. Hence, the signature strength of such objects is in the same order as terrain RCS in the interesting incidence angle range of around 20–60°. This circumstance also makes clear that lower incidence angles are not useful due to strongly increasing background clutter, and higher incidence angles are less useful since the penetration capability into the ground decreases, and due to longer ranges, the signal-to-noise ratio becomes unacceptable. However, even in the useful range of incidence angles the discrimination of target signatures from clutter can be a serious problem.
3 Philosophy and Design of the Radar System
Following previous considerations, a SAR system using multiple channels, different polarization combinations, and operating in the UHF range was developed. Next the main features and their main drivers are briefly outlined:
3.1 SAR Principle and Spatial Resolution
SAR is an imaging principle based on a side-looking geometry, hence the radar can move on safe ground, while imaging the hazardous ground. Perpendicular to the linear radar motion, i.e., across track, narrow radar pulses are used to scan the area of interest. The delays of the pulses correspond to the range in across track direction. The width of the pulses, i.e., the bandwidth of the used signal, determines the resolution in time and consequently the range resolution via the knowledge of the speed of light. In the direction of radar motion, i.e., along track, the spatial resolution is achieved by coherently superimposing the radar echoes for a certain path length, called the synthetic aperture. By that way a large antenna is synthesized by the radar motion, similar to a real-aperture antenna of that size being not useful in practice due its size and mass. The size of the image is determined by the minimum and maximum across track distance to be scanned, and the length of the synthetic aperture.
3.2 UHF Range
Due to penetration depth constraints the frequency range should be lower than a few GHz. To use reasonable antenna sizes, the lowest frequency should be at least in the order of few 100 MHz.
3.3 Multiple Channels
The constraints of very limited target-to-clutter ratios to be expected for typical landmine/UXO scenarios require methods to improve that situation. From experience it was expected that a man-made target as a landmine typically shows in average similar signature strengths, even when incidence angle, bistatic angle, and aspect angle of observation are changed considerably. The required variability of angles can be achieved by high-resolution SAR, i.e., a large range of aspect angles for producing the synthetic aperture, and the use of multiple transmitters and receivers located on different positions in elevation. The latter produces on one hand different incidence angles for each TX/RX combination, and on the other hand, different bistatic angles when the average distance to the area of interest is sufficiently close to the radar. If furthermore the antenna arrangement for TX and RX is made such that the antenna array produces certain length in vertical direction, additionally vertical spatial resolution can be achieved, thus allowing a kind of three-dimensional imaging. In contrast to man-made targets, the natural background clutter usually is composed of multiple reflections within a resolution cell. Hence, their coherent superposition within the resolution cell changes quickly with different observation angles. In summary, by superposition of radar images from different observation angles the clutter is averaging out more than the man-made objects, producing a net gain in target-to-clutter ratio.
4 Representative Measurement Results
Now the buried targets can be clearly discriminated from the clutter. While targets 2 and 7 are detected for HH polarization they are not for VV polarization. Only target 4 is missing in both images, probably due to the fact that the mine was only simulated by an empty case for which the wall thickness is quite small compared to the wavelengths. However, this result shows the importance and necessity of precisely superimposing several SAR images of different observation angles and using polarimetric information. This statement could be verified by various other experiments showing similar results.
A novel radar approach has been developed and applied to the problem of reliable landmine and UXO detection at high throughput of scanned area per time. This method allows detection and localization of buried and surface objects in the presence of background clutter, while up to now no discrimination of objects concerning a threat or a false alarm is performed. Moreover, the philosophy of the concept is to perform first reliable and effective detection of suspicious objects, and then to use other sensors based on different physical principles to investigate further those locations where detections occurred. By that way the overall system acts as a multi-sensor system, being the only successful approach to face and solve the problem for our opinion.
The use of radar for buried object detection is based on the measurement of bulk material anomalies, producing mismatch of wave impedance, and thus giving rise for wave scattering and reflections. Microwave radar thus is not sensitive to spectroscopic properties, and thus cannot discriminate explosives from other matter just by signature. Consequently, the multi-sensor approach as mentioned above will do the discrimination.
The use of SAR technology provides the unique benefits of combining side-looking geometry for operation from safe area into hazardous area, high area throughput due to large image size, and highest spatial resolution for detection of even small and weak objects in RCS. The problem of background clutter is solved by applying a multi-channel system in TX and RX, providing different observation angles for each TX-RX combination. In addition, the use of wave polarization improves further considerably detection performance.
Although TIRAMI-SAR already performs satisfying at this stage, it offers plenty of room for further optimization and improvements. Especially, the use of other antenna array geometries, other waveforms, other polarization combinations, etc., is of major interest. Furthermore, many other experiments and field trials have to be executed to make performance assessment more reliable and informative.
Funding was provided by EU Framework Programme 7 (FP7-SEC-2011-1 No. 284747).
- Feng X et al (2005) Estimation of ground surface topography and velocity model by SARGPR and its application to landmine detection. In: SPIE conference detection and remediation for mines and minelike targets X, Vol. 5794Google Scholar
- Globalgiving (2016). https://www.globalgiving.org/projects/mli-mine-detection-dogs/. Accessed 27 July 2016
- Heinzel A et al (2016) Focusing methods for ground penetrating MIMO SAR Imaging within half-spaces of different permittivity. In: Proceedings of 11th European conference on synthetic aperture radar EUSAR’16, pp 842–846Google Scholar
- Kabourek V et al (2012) Clutter reduction based on principal component analysis technique for hidden objects detection. Radioengineering 21(1):464–470Google Scholar
- Landmine Monitor (2015). http://www.the-monitor.org/en-gb/reports/2015/landmine-monitor-2015/preface.aspx. Accessed 27 July 2016
- Liu G et al (2003) SAR imaging for a forward-looking GPR system, In: SPIE conference detection and remediation for mines and minelike targets VIII, Vol. 5089Google Scholar
- Lloyd D, Longstaff I D (2003) Ultra-wideband multistatic SAR for the detection and location of landmines. IEE proceedings. Online No. 20030444Google Scholar
- Pica.Army (2016). http://www.pica.army.mil/Picatinny/products_services/products15.aspx. Accessed 27 July 2016
- Schreiber E et al (2016) Challenges for operational use of ground-based high-resolution SAR for landmines and UXO detection. In: Proceedings of 11th European conference on synthetic aperture radar EUSAR’16, pp 834–837Google Scholar
- Telovation (2016). http://www.telovation.com/articles/land-mine-detection-robot.html. Accessed 27 July 2016
- Theleadsouthaustralia (2016). http://www.theleadsouthaustralia.com.au/industries/defence/minelab-to-supply-us-military-with-mounted-mine-detection-system/. Accessed 27 July 2016
- TIRAMISU Project (2012) Project website. http://fp7-tiramisu.eu. Accessed 27 July 2016
- Ulaby FT (1981) Microwave remote sensing—active and passive. Microwave remote sensing fundamentals and radiometry, vol I. Addison-Wesley, ReadingGoogle Scholar
- Ulaby FT (1982) Microwave remote sensing—active and passive. Radar remote sensing and surface and emission theory, vol II. Addison-Wesley, ReadingGoogle Scholar
- Ulaby FT (1989) Handbook of radar scattering statistics for terrain. Artech House, NorwoodGoogle Scholar
- Ulaby FT (1992) Modelling radar backscatter from vegetation. Short course notes. DLR, OberpfaffenhofenGoogle Scholar
- Wu H et al (2010) Motion compensation for landmine detecting vehicle-borne SAR. In: Proceedings of IEEE 11th international radar symposium IRS’10Google Scholar