This PDF file contains the front matter associated with SPIE Proceedings Volume 11418, including the Title Page, Copyright Information, and Table of Contents.

The focus of this article is deep learning on small, class imbalanced data sets in support of explosive hazard detection (EHD) and automatic target recognition (ATR). To this end, we explore artificial neural networks that are driven by similarity versus classification or regression. Similarity can be emphasized via network design, e.g., siamese networks, and/or underlying metric, e.g., contrastive or triple loss. The general goal of a similarity neural network (SNN) is discriminative training via focusing on similarity between tuples of like (and unlike) inputs. As such, SNNs have the potential to learn improved solutions on small data; aka do more with less". Herein, we explore different avenues and we show that SNNs are essentially neural feature extractors followed by k-nearest neighbor classification. Instead of experimenting on a government data set that cannot be shared, we instead focus on benchmark community data sets for sake of reproducible research. Preliminary findings are encouraging as they suggest that SNNs have great potential for tasks like EHD and ATR.

The focus of this article is extending classifiers from N classes to N+1 classes without retraining for tasks like explosive hazard detection (EHD) and automatic target recognition (ATR). In recent years, deep learning has become state-of-the-art across domains. However, algorithms like convolutional neural networks (CNNs) suffer from the assumption of a closed-world model. That is, once a model is learned, a new class cannot usually be added without changes in the architecture and retraining. Herein, we put forth a way to extend a number of deep learning algorithms while keeping their features in a locked state; i.e., features are not retrained for the new N+1 class. Different feature transformations, metrics, and classifiers are explored to assess the degree to which a new sample belongs to one of the N classes and a decision rule is used for classification. Whereas this extends a deep learner, it does not tell us if a network with locked features has the potential to be extended. Therefore, we put forth a new method based on visually assessing cluster tendency to assess the degree to which a deep learner can be extended (or not). Lastly, while we are primarily focused on tasks like aerial EHD and ATR, experiments herein are for benchmark community data sets for sake of reproducible research.

Ground Penetrating Radar (GPR) is a geophysical method developed for subsurface imaging. The main components of a GPR system are the antenna, the control unit, and the encoder. The GPR system moves along the area of interest and a set of data, called radargram, is obtained. In a single-short offset GPR acquisition, the distance between the transmitter and receiver antennas is very short. The collected data over an area can be used to construct images of the dielectric permittivity and the electric conductivity of the subsurface, which allows identifying objects such as anti-personnel mines, pipes, people in an earthquake, fossils, among others. Full waveform inversion (FWI) of GPR data can be used to estimate such electromagnetic parameters based on the minimization of a misfit function. However, FWI of GPR data has several problems: cross-talk between parameters, the estimation of the amplitude and signature of the electromagnetic source, the presence of noise in the data, the non-uniqueness of the solution and the high computational cost. In this paper, an alternative cost function for FWI is proposed to mitigate the problem of the estimation of the amplitude of the electromagnetic source. We show that the proposed alternative cost function for FWI is more robust than traditional norm in noisy environments. The proposed cost function is tested over a synthetic model with complex structures called SEAM and over collected data in a region of Colombia. In order to deal with the high computational complexity of FWI, all the experiments are computed in a cluster of GPUs.

In this study, we present a brief information about GPR data processing and then processing methods are proposed dedicated to false alarm reduction with varying antenna height. We used horizontal filtering in cross-track and continuous wavelet transformation for A-scan signals. Additionally, 2D Wavelets and Gabor filters are applied to the data. Comparative results are presented over real data set obtained from various buried objects. It is observed that the horizontal filtering gives satisfactory results especially in cases where there is variability in antenna height.

Two studies were conducted at 43° 43' 24" N, 72° 16' 24" W during which the thermal signature of an area containing buried objects was observed. The two periods were from September – November 2016 and May – December 2018. In the first phase, the soil was left in its natural state while in the second, it was removed to a depth of 65 cm, homogenized, then replaced. In both instances, the physical properties of the soil were fully characterized, the meteorological forcing recorded, and subsurface moisture and temperature states measured. In phase one four objects, two round and two rectangular, were buried so that their tops were at a depth of 5 cm. In the second experiment four rectangular objects were emplaced at depths of 5 and 25 cm. In both phases two of the objects were metal and two plastic. Differences in thermal signatures of the buried objects will be discussed as they relate to emplacement depth, soil properties, and object composition.

Artificial intelligence and machine learning algorithms for object detection in the infrared require an extensive amount of high-resolution object-tagged thermal infrared images. Often, acquiring real imagery of sufficient size and range of environmental conditions is difficult due to the cost and time. To address this need, the current study has developed a novel computational framework, i.e. the Sensor Engine, that generates target-tagged synthetic infrared imagery of large complex natural environments. This computational framework, coupled with high-fidelity soil and vegetation thermal physics and geometry models, generates synthetic, high-definition infrared images tailored for High-Performance Computing (HPC) systems. A unique plugin mechanism used to load and unload configured infrared sensors at run time in addition to allowing the framework to effectively work with different sensors in parallel is also discussed. The sensor model within the Sensor Engine communicates with another computational framework to acquire radiative energy for each sensor pixel detector as well as material, distance, source location, and incident angle. To demonstrate the modus operandi of this computational framework, an evaluation and discussion of runtime message passing and test cases are provided.

Distributed acoustic sensing (DAS) systems using fiber optic (FO) cables are becoming commonplace in many perimeter security applications. They have the advantage of cost-effectively covering large geographic expanses without temporal or spatial gaps. Recent technology advances in commercial DAS systems have significantly improved system sensitivity and the ability to generate time-harmonic waveforms similar to microphones and geophones. These waveforms can be saved at regular channel intervals along the length of the cable. Relative to microphones and geophones, there are a number of DAS limitations, including generally lower system sensitivity, random signal fading,¹ and strong longitudinal radiation pattern effects. The latter limits DAS systems for detecting seismic-acoustic waves propagating perpendicular to the cable. In this paper, we investigate the use of coiled bundles of FO cable. We show that coherently stacking channel waveforms in coils improves signal-to-noise ratio (SNR) by a root mean square (RMS) factor of 6 relative to noise in unstacked channels, and that the coils generate a more omnidirectional radiation pattern relative to straight cable segments. However, this is at the expense of decreased signal power and more complex installation methods. These developments are incremental steps in enhancing our ability to use DAS FO systems for tracking ground and air vehicles. In our test, we used a commercially available FO DAS system configured to monitor ground vehicles and low-flying aircraft. The experiment was conducted in the deep sandy soils of the New Jersey Pine Barrens.² The two primary DAS components were a laser interrogator unit (IU) and acoustically sensitive FO cable (both manufactured by OptaSense). For redundancy, we used an additional IU from the Naval Research Laboratory. We buried 3,000 m of FO cable at an approximate 30 cm depth, collecting 2,900 channels of strain time series data. We constructed four arrays by coiling cable in discrete bundles. A single bundle could contain 16-24 m of cable wound on a 20 cm diameter jig. Across the entire emplacement, we made roughly 100 such bundles (i.e. 2/3 of the buried cable was wound into coils). As a result, the sensor covered a straight-line distance of approximately 1,000 m. This paper focuses exclusively on the FO array design and installation, as well as the processing methods and benefits of using coiled arrays. Our results indicate that these methods have significant merit for enhancing DAS air and ground vehicle detection and tracking

This paper focuses results obtained from a field spectroscopy campaigns for detecting underground structures. A number of vegetation indices such as the Normalized Difference Vegetation Index (NDVI), Simple Ratio (SR), Enhanced Vegetation Index (EVI) and Ratio Vegetation Index (RVI) were utilized for the development of a vegetation index-based procedure aiming at the detection of underground military structures by using existing vegetation indices. The measurements were taken at the following test areas such as: (a) vegetation area covered with the vegetation (barley), in the presence of an underground military structure (b) vegetation area covered with the vegetation (barley), in the absence of an underground military structure. For this purpose was using histograms to obtain useful information about Vegetation Indices spectral behaviours and to compare the two testing areas.

The purpose of this paper is to present the results obtained from field spectroscopy campaigns for detecting underground structures. Simple Ratio (SR) Vegetation Index, was utilized for the development of a Vegetation Index-based procedure aiming at the detection of underground military. The measurements were taken at the following test areas such as: Area (a) vegetation area covered with the vegetation (barley), in the presence of an underground military structure and Area (b) vegetation area covered with the vegetation (barley), in the absence of an underground military structure. The results show that the SR Vegetation Index is highly useful in detecting deep man-made infrastructures in Cyprus.

The "EXCELSIOR" H2020 Widespread Teaming Phase 2 Project: ERATOSTHENES: EXcellence Research Centre for Earth SurveiLlance and Space-Based MonItoring Of the EnviRonment is supported from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 857510 for a 7 year project period to establish a Centre of Excellence in Cyprus. As well, the Government of the Republic of Cyprus is providing additional resources to support the establishment of the ERATOSTHENES Centre of Excellence (ECoE) in Cyprus. The ECoE seeks to fill the gap by assisting in the spaceborne Earth Observation activities in the Eastern Mediterranean and become a regional key player in the Earth Observation (EO) sector. There are distinct needs and opportunities that motivate the establishment of an Earth Observation Centre of Excellence in Cyprus, which are primarily related to the geostrategic location of the European Union member state of Cyprus to examine complex scientific problems and address user needs in the Eastern Mediterranean, Middle East and Northern Africa (EMMENA), as well as South-East Europe. An important objective of the ECoE is to be a Digital Innovation Hub and a Research Excellence Centre for EO in the EMMENA region, which will establish an ecosystem where state-of-the-art sensing technology, cutting-edge research, targeted education services, and entrepreneurship come together. It is based on the paradigm of Open Innovation 2.0 (OI2.0), which is founded on the Quadruple Helix Model, where Government, Industry, Academia and Society work together to drive change by taking full advantage of the cross-fertilization of ideas. The ECoE as a Digital Innovation Hub (DIH) adopts a two-axis model, where the vertical axis consists of three Thematic Clusters for sustained excellence in research of the ECoE in the domains of Atmosphere and Climate, Resilient Societies and Big Earth Data Management, while the horizontal axis is built around four functional areas, namely: Infrastructure, Research, Education, and Entrepreneurship. The ECoE will focus on five application areas, which include Climate Change Monitoring, Water Resource Management, Disaster Risk Reduction, Access to Energy and Big EO Data Analytics. This structure is expected to leverage the existing regional capacities and advance the excellence by creating new programs and research, thereby establishing the ECoE as a worldclass centre capable of enabling innovation and research competence in Earth Observation, actively participating in Europe, the EMMENA region and the global Earth Observation arena. The partners of the EXCELSIOR consortium include the Cyprus University of Technology as the Coordinator, the German Aerospace Center (DLR), the Leibniz Institute for Tropospheric Research (TROPOS), the National Observatory of Athens (NOA) and the Department of Electronic Communications, Deputy Ministry of Research, Innovation and Digital Policy.

Modern landmine detection capability struggles to discriminate between dangerous targets and harmless clutter, which increases the risk and time required for mine-clearing operations. To reduce risk to the warfighter, Vadum is developing the Vibration-ENhanced Underground Sensing (VENUS) system for improved landmine identification capability. The novel sensor uses an electromagnetic stimulus to induce mechanical vibrations in buried targets that are detectable using a sensitive RF vibrometer. Landmines produce unique phenomenological responses to the stimulus as a result of specific structural features that distinguish targets from other metal clutter. This physics-based approach enables discrimination even for low-metal-content (LMC) antipersonnel mines. In this paper, a high-level system architecture and test results from the VENUS hardware prototype are presented. A high dynamic range RF vibrometer design enables extraction of small vibration-modulated signals less than 200 Hz from the RF carrier. Distinct vibrational responses collected by the hardware prototype are shown for several buried inert antipersonnel landmines to demonstrate feasibility at up to two inches of burial depth. System capability tradeoffs are discussed and compared to the state-of-the-art including sensitivity, detection depth, discrimination capability, and achievable size, weight, and power (SWAP).

Recently, a novel optical technology, LIBS in mid-IR (MWIR/LWIR) region was developed to capture the infrared molecular emission signatures from those vibrationally excited intact sample molecules excited by laser-induced plasma. Mid-IR LIBS is the first mid-IR emission spectroscopy that can complement LIBS and Raman as rapid, in situ, and standoff chemical characterization probes without the need of any sample preparation. With all the advantages of the conventional UVN LIBS, the UVN + LWIR LIBS spectrometer can rapidly and unambiguously reveal both the elemental composition and molecular makeups of the sample that is meters away without any sample preparation required and without the need to unscramble the spectral fingerprints of targets from the irregular and cluttered background. UVN + LWIR LIBS is able to provide in-situ, real-time/near-real-time chemical detection and identification regardless of the shapes and conditions of the sample surface while requiring no need for any sample preparations. One does not need to do anything to the target sample, just point the laser at the intended target and get the spectral signatures back within a millisecond.

Maritime surveillance is of critical importance for threat prevention and maintenance of national security and safety. Maritime traffic comprises worldwide navigation of millions of vessels. In this sense, the geostrategic position of Cyprus entails the need for effective monitoring of marine traffic. Remote Sensing is a technique, which enables maritime surveillance by means of space-based detection and identification of marine traffic. The advent of new satellite missions, such as Sentinel-1, enabled the acquisition of systematic datasets for monitoring vessels. Using the Copernicus Open Access Hub service, it is now feasible to access satellite data in a fully automated and near real-time mode and deliver vessel information through a web portal interface. Nevertheless, there is still a great need to understand the full potential of the information acquired from such sensors. In this paper, an overview of vessel tracking techniques using Sentinel acquisitions is carried out. Consequently, vessel detections via space imagery could be authenticated against Automatic Identification System (AIS) data, which provide the location and dimensions of ships that are legally operating in the Cyprus region.

Publisher’s Note: This paper, originally published on 24 April 2020, was replaced with a corrected/revised version on 8 June 2020. If you downloaded the original PDF but are unable to access the revision, please contact SPIE Digital Library Customer Service for assistance.