Deep neural networks are empirically derived systems that have transformed research methods and are driving scientific discovery. Artificial electromagnetic materials, such as electromagnetic metamaterials, photonic crystals, and plasmonics, are research fields where deep neural network results evince the data driven approach; especially in cases where conventional computational and optimization methods have failed. We propose and demonstrate a deep learning method capable of finding accurate solutions to ill-posed inverse problems, where the conditions of existence and uniqueness are violated. A specific example of finding the metasurface geometry which yields a radiant exitance matching the external quantum efficiency of gallium antimonide is demonstrated.
In this work we consider the problem of developing deep learning models – such as convolutional neural networks (CNNs) - for automatic target detection (ATD) in infrared (IR) imagery. CNN-based ATD systems must be trained to recognize objects using bounding box (BB) annotations generated by human annotators. We hypothesize that individual annotators may exhibit different biases and/or variability in the characteristics of their BB annotations. Similarly, computer-aided annotation methods may also introduce different types of variability into the BBs. In this work we investigate the impact of BB variability on the behavior and detection performance of CNNs trained using them. We consider two specific BB characteristics here: the center-point, and the overall scale of BBs (with respect to the visual extent of the targets they label). We systematically vary the bias or variance of these characteristics within a large training dataset of IR imagery, and then evaluate the performance on the resulting trained CNN models. Our results indicate that biases in these BB characteristics do not impact performance, but will cause the CNN to mirror the biases in its BB predictions. In contrast, variance in these BB characteristics substantially degrades performance, suggesting care should be taken to reduce variance in the BBs.
In this work we consider the problem of developing algorithms for the automatic detection of buried threats in handheld Ground Penetrating Radar (HH-GPR) data. The development of algorithms for HH-GPR is relatively nascent compared to larger downward-looking GPR (DL-GPR) systems. A large number of buried threat detection (BTD) algorithms have been developed for DL-GPR systems. Given the similarities between DL-GPR data and HHGPR data, effective BTD algorithm designs may be similar for both modalities. In this work we explore the application of successful class of DL-GPR-based algorithms to HH-GPR data. In particular, we consider the class of algorithms that are based upon gradient-based features, such as histogram-of-oriented gradients (HOG) and edge histogram descriptors. We apply a generic gradient-based feature with a support vector machine to a large dataset of HH-GPR data with known buried threat locations. We measure the detection performance of the algorithm as we vary several important design parameters of the feature, and identify those designs that yield the best performance. The results suggest that the design of the gradient histogram (GH) feature has a substantial impact on its performance. We find that a tuned GH algorithm yields substantially-better performance, but ultimately performs similarly to the energy-based detector. This suggests that GH-based features may not be beneficial for HH-GPR data, or that further innovation will be needed to achieve benefits.
KEYWORDS: General packet radio service, Convolutional neural networks, Detection and tracking algorithms, Neurons, Control systems, Performance modeling, Data processing, Algorithm development, Data modeling, Network architectures
The ground penetrating radar (GPR) is a remote sensing technology that has been successfully used for detecting buried explosive threats. A large body of published research has focused on developing algorithms that automatically detect buried threats using data from GPR sensors. One promising class of algorithms for this purpose is convolutional neural networks (CNNs), however CNNs suffer from overfitting due to the limited and variable nature of GPR data. One solution to this problem is to use a validation dataset during training, however this excludes valuable labeled data from training. In this work we show that two modern techniques for training CNNs – Batch Normalization and the Adam Optimizer - substantially improve CNN performance and reduce overfitting when applied jointly. We also investigate and identify useful settings for several important CNN hyperparameters: l2 regularization, Dropout, and the learning rate schedule. We find that the improved CNN (a baseline CNN, plus all of our improvements) substantially outperforms two competing conventional detection algorithms.
KEYWORDS: Algorithm development, General packet radio service, Detection and tracking algorithms, Sensors, Ground penetrating radar, Antennas, Feature extraction, Data modeling, Radar, Systems modeling
In this work we consider the problem of developing algorithms for the automatic detection of buried threats using handheld Ground Penetrating Radar (HH-GPR) data. The development of algorithms for HH-GPR is relatively nascent compared to algorithm development efforts for larger downward-looking GPR (DL-GPR) systems. One of the biggest bottlenecks for the development of algorithms is the relative scarcity of labeled HH-GPR data that can be used for development. Given the similarities between DL-GPR data and HH-GPR data however, we hypothesized that it may be possible to utilize DL-GPR data to support the development of algorithms for HH-GPR. In this work we assess the detection performance of a HH-GPR-based BTD algorithm as we vary the amounts and characteristics of the DL-GPR data included in the development of HH-GPR detection algorithms. The results indicate that supplementing HH-GPR data with DL-GPR does improve performance, especially when including data collected over buried threat locations.
KEYWORDS: Detection and tracking algorithms, Data modeling, Ground penetrating radar, Algorithm development, Data processing, Feature extraction, Visual process modeling, Unexploded object detection, Threat warning systems
A great deal of research has been focused on the development of computer algorithms for buried threat detection (BTD) in ground penetrating radar (GPR) data. Most recently proposed BTD algorithms are supervised, and therefore they employ machine learning models that infer their parameters using training data. Cross-validation (CV) is a popular method for evaluating the performance of such algorithms, in which the available data is systematically split into ܰ disjoint subsets, and an algorithm is repeatedly trained on ܰ−1 subsets and tested on the excluded subset. There are several common types of CV in BTD, which vary principally upon the spatial criterion used to partition the data: site-based, lane-based, region-based, etc. The performance metrics obtained via CV are often used to suggest the superiority of one model over others, however, most studies utilize just one type of CV, and the impact of this choice is unclear. Here we employ several types of CV to evaluate algorithms from a recent large-scale BTD study. The results indicate that the rank-order of the performance of the algorithms varies substantially depending upon which type of CV is used. For example, the rank-1 algorithm for region-based CV is the lowest ranked algorithm for site-based CV. This suggests that any algorithm results should be interpreted carefully with respect to the type of CV employed. We discuss some potential interpretations of performance, given a particular type of CV.
A large number of algorithms have been proposed for automatic buried threat detection (BTD) in ground penetrating radar (GPR) data. Convolutional neural networks (CNNs) have recently achieved groundbreaking results on many recognition tasks. This success is due, in part, to their ability to automatically infer effective data representations (i.e., features) using training data. This capability however results in a high capacity model (i.e., many free parameters) that is difficult to train, and more prone to overfitting, than models employing hand-crafted feature designs. This drawback is pronounced when training data is relatively scarce, as is the case with GPR BTD. In this work we propose to combine the relative advantages of hand-crafted features, and CNNs, by constructing CNN architectures that closely emulate successful hand-crafted feature designs for GPR BTD. This makes it possible to apply supervised training to traditional hand-crafted features, allowing them to adapt to the unique characteristics of the GPR BTD problem. Simultaneously, this approach yields a much lower capacity CNN model that incorporates substantial prior research knowledge, making the model much easier to train. We demonstrate the feasibility and effectiveness of this approach by designing a “neural” implementation of the popular histogram of oriented gradient (HOG) feature. The resulting neural HOG (NHOG) implementation is much smaller and easier to train than standard CNN architectures, and achieves superior detection performance compared to the un-trained HOG feature. In theory, neural implementations can be developed for many existing successful GPR BTD algorithms, potentially yielding similar benefits.
In this work, we consider the development of algorithms for automated buried threat detection (BTD) using Ground Penetrating Radar (GPR) data. When viewed in GPR imagery, buried threats often exhibit hyperbolic shapes, and this characteristic shape can be leveraged for buried threat detection. Consequentially, many modern detectors initiate processing the received data by extracting visual descriptors of the GPR data (i.e., features). Ideally, these descriptors succinctly encode all decision-relevant information, such as shape, while suppressing spurious data content (e.g., random noise). Some notable examples of successful descriptors include the histogram of oriented gradient (HOG), and the edge histogram descriptor (EHD). A key difference between many descriptors is the precision with which shape information is encoded. For example, HOG encodes shape variations over both space and time (high precision); while EHD primarily encodes shape variations only over space (lower precision). In this work, we conduct experiments on a large GPR dataset that suggest EHD-like descriptors outperform HOG-like descriptors, as well as exhibiting several other practical advantages. These results suggest that higher resolution shape information (particularly shape variations over time) is not beneficial for buried threat detection. Subsequent analysis also indicates that the performance advantage of EHD is most pronounced among difficult buried threats, which also exhibit more irregular shape patterns.
KEYWORDS: General packet radio service, Machine learning, Data analysis, Algorithm development, Detection and tracking algorithms, Data modeling, Ground penetrating radar, Visualization, Antennas, Feature extraction
This work focuses on the development of automatic buried threat detection (BTD) algorithms using ground penetrating radar (GPR) data. Buried threats tend to exhibit unique characteristics in GPR imagery, such as high energy hyperbolic shapes, which can be leveraged for detection. Many recent BTD algorithms are supervised, and therefore they require training with exemplars of GPR data collected over non-threat locations and threat locations, respectively. Frequently, data from non-threat GPR examples will exhibit high energy hyperbolic patterns, similar to those observed from a buried threat. Is it still useful therefore, to include such examples during algorithm training, and encourage an algorithm to label such data as a non-threat? Similarly, some true buried threat examples exhibit very little distinctive threat-like patterns. We investigate whether it is beneficial to treat such GPR data examples as mislabeled, and either (i) relabel them, or (ii) remove them from training. We study this problem using two algorithms to automatically identify mislabeled examples, if they are present, and examine the impact of removing or relabeling them for training. We conduct these experiments on a large collection of GPR data with several state-of-the-art GPR-based BTD algorithms.
The Ground Penetrating Radar (GPR) is a remote sensing modality that has been used to collect data for the task of buried threat detection. The returns of the GPR can be organized as images in which the characteristic visual patterns of threats can be leveraged for detection using visual descriptors. Recently, convolutional neural networks (CNNs) have been applied to this problem, inspired by their state-of-the-art-performance on object recognition tasks in natural images. One well known limitation of CNNs is that they require large amounts of data for training (i.e., parameter inference) to avoid overfitting (i.e., poor generalization). This presents a major challenge for target detection in GPR because of the (relatively) few labeled examples of targets and non-target GPR data. In this work we use a popular transfer learning approach for CNNs to address this problem. In this approach we train two CNN on other, much larger, datasets of grayscale imagery for different problems. Specifically, we pre-train our CNNs on (i) the popular Cifar10 dataset, and (ii) a dataset of high resolution aerial imagery for detecting solar photovoltaic arrays. We then use varying subsets of the parameters from these two pre-trained CNNs to initialize the training of our buried threat detection networks for GPR data. We conduct experiments on a large collection of GPR data and demonstrate that these approaches improve the performance of CNNs for buried target detection in GPR data
Ground penetrating radar (GPR) systems have emerged as a state-of-the-art remote sensing platform for the automatic detection of buried explosive threats. The GPR system that was used to collect the data considered in this work consists of an array of radar antennas mounted on the front of a vehicle. The GPR data is collected as the vehicle moves forward down a road, lane or path. The data is then processed by computerized algorithms that are designed to automatically detect the presence of buried threats. The amount of GPR data collected is typically prohibitive for real-time buried threat detection and therefore it is common practice to first apply a prescreening algorithm in order to identify a small subset of data that will then be processed by more computationally advanced algorithms. Historically, the F1V4 anomaly detector, which is energy-based, has been used as the prescreener for the GPR system considered in this work. Because F1V4 is energy-based, it largely discards shape information, however shape information has been established as an important cue for the presence of a buried threat. One recently developed prescreener, termed the HOG prescreener, employs a Histogram of Oriented Gradients (HOG) descriptor to leverage both energy and shape information for prescreening. To date, the HOG prescreener yielded inferior performance compared to F1V4, even though it leveraged the addition of shape information. In this work we propose several modifications to the original HOG prescreener and use a large collection of GPR data to demonstrate its superior detection performance compared to the original HOG prescreener, as well as to the F1V4 prescreener.
Forward-looking ground penetrating radar (FLGPR) is a remote sensing modality that has been investigated for buried threat detection. The FLGPR considered in this work consists of a sensor array mounted on the front of a vehicle, which inspects an area in front of the vehicle as it moves down a lane. The FLGPR collects data using a stepped frequency approach, and the received radar data is processed by filtered backprojection to create images of the subsurface. A large body of research has focused on developing effective supervised machine learning algorithms to automatically discriminate between imagery associated with target and non-target FLGPR responses. An important component of these automated algorithms is the design of effective features (e.g., image descriptors) that are extracted from the FLGPR imagery and then provided to the machine learning classifiers (e.g., support vector machines). One feature that has recently been proposed is computed from the magnitude of the two-dimensional fast Fourier transform (2DFFT) of the FLGPR imagery. This paper presents a modified version of the 2DFFT feature, termed 2DFFT+, that yields substantial detection performance when compared with several other existing features on a large collection of FLGPR imagery. Further, we show that using partial least-squares discriminative dimensionality reduction, it is possible to dramatically lower the dimensionality of the 2DFFT+ feature from 2652 dimensions down to twenty dimensions (on average), while simultaneously improving its performance.
In recent years, the Ground Penetrating Radar (GPR) has successfully been applied to the problem of buried threat detection (BTD). A large body of research has focused on using computerized algorithms to automatically discriminate between buried threats and subsurface clutter in GPR data. For this purpose, the GPR data is frequently treated as an image of the subsurface, within which the reflections associated with targets often appear with a characteristic shape. In recent years, shape descriptors from the natural image processing literature have been applied to buried threat detection, and the histogram of oriented gradient (HOG) feature has achieved state-of-the-art performance. HOG consists of computing histograms of the image gradients in disjoint square regions, which we call pooling regions, across the GPR images. In this work we create a large body of potential pooling regions and use the group LASSO (GLASSO) to choose a subset of the pooling regions that are most appropriate for BTD on GPR data. We examined this approach on a large collection of GPR data using lane-based cross-validation, and the results indicate that GLASSO can select a subset of pooling regions that lead to superior performance to the original HOG feature, while simultaneously also reducing the total number of features needed. The selected pooling regions also provide insight about the regions in GPR images that are most important for discriminating threat and nonthreat data.
The ground penetrating radar (GPR) is a popular remote sensing modality for buried threat detection. In this work we focus on the development of supervised machine learning algorithms that automatically identify buried threats in GPR data. An important step in many of these algorithms is feature extraction, where statistics or other measures are computed from the raw GPR data, and then provided to the machine learning algorithms for classification. It is well known that an effective feature can lead to major performance improvements and, as a result, a variety of features have been proposed in the literature. Most of these features have been handcrafted, or designed through trial and error experimentation. Dictionary learning is a class of algorithms that attempt to automatically learn effective features directly from the data (e.g., raw GPR data), with little or no supervision. Dictionary learning methods have yielded state-of-theart performance on many problems, including image recognition, and in this work we adapt them to GPR data in order to learn effective features for buried threat classification. We employ the LC-KSVD algorithm, which is a discriminative dictionary learning approach, as opposed to a purely reconstructive one like the popular K-SVD algorithm. We use a large collection of GPR data to show that LC-KSVD outperforms two other approaches: the popular Histogram of oriented gradient (HOG) with a linear classifier, and HOG with a nonlinear classifier (the Random Forest).
KEYWORDS: Systems modeling, Forward looking infrared, Target detection, Sensors, Performance modeling, General packet radio service, Detection and tracking algorithms, Data modeling, Sensor performance, Computing systems
Many remote sensing modalities have been developed for buried target detection (BTD), each one offering
relative advantages over the others. There has been interest in combining several modalities into a single BTD system
that benefits from the advantages of each constituent sensor. Recently an approach was developed, called multi-state
management (MSM), that aims to achieve this goal by separating BTD system operation into discrete states, each with
different sensor activity and system velocity. Additionally, a modeling approach, called Q-MSM, was developed to
quickly analyze multi-modality BTD systems operating with MSM. This work extends previous work by demonstrating
how Q-MSM modeling can be used to design BTD systems operating with MSM, and to guide research to yield the most
performance benefits. In this work an MSM system is considered that combines a forward-looking infrared (FLIR)
camera and a ground penetrating radar (GPR). Experiments are conducted using a dataset of real, field-collected, data
which demonstrates how the Q-MSM model can be used to evaluate performance benefits of altering, or improving via
research investment, various characteristics of the GPR and FLIR systems. Q-MSM permits fast analysis that can
determine where system improvements will have the greatest impact, and can therefore help guide BTD research.
KEYWORDS: Target detection, Ground penetrating radar, Algorithm development, Remote sensing, Detection and tracking algorithms, General packet radio service, Sensors, Data processing, Detector development, Signal attenuation
Ground penetrating radar (GPR) is a popular remote sensing modality for buried threat detection. Many algorithms
have been developed to detect buried threats using GPR data. One on-going challenge with GPR is the detection of
very deeply buried targets. In this work a detection approach is proposed that improves the detection of very deeply
buried targets, and interestingly, shallow targets as well. First, it is shown that the signal of a target (the target
“signature”) is well localized in time, and well correlated with the target’s burial depth. This motivates the proposed
approach, where GPR data is split into two disjoint subsets: an early and late portion corresponding to the time at
which shallow and deep target signatures appear, respectively. Experiments are conducted on real GPR data using
the previously published histogram of oriented gradients (HOG) prescreener: a fast supervised processing method
operated on HOG features. The results show substantial improvements in detection of very deeply buried targets
(4.1% to 17.2%) and in overall detection performance (81.1% to 83.9%). Further, it is shown that the performance
of the proposed approach is relatively insensitive to the time at which the data is split. These results suggest that
other detection methods may benefit from depth-based processing as well.
The forward-looking ground penetrating radar (FLGPR) is a remote sensing modality that has recently been investigated
for buried threat detection. The FLGPR considered in this work uses stepped frequency sensing followed by filtered
backprojection to create images of the ground, where each image pixel corresponds to the radar energy reflected from
the subsurface at that location. Typical target detection processing begins with a prescreening operation where a small
subset of spatial locations are chosen to consider for further processing. Image statistics, or features, are then extracted
around each selected location and used for training a machine learning classification algorithm. A variety of features
have been proposed in the literature for use in classification. Thus far, however, predominantly hand-crafted or
manually designed features from the computer vision literature have been employed (e.g., HOG, Gabor filtering, etc.).
Recently, it has been shown that image features learned directly from data can obtain state-of-the-art performance on a
variety of problems. In this work we employ a feature learning scheme using k-means and a bag-of-visual-words model
to learn effective features for target and non-target discrimination in FLGPR data. Experiments are conducted using
several lanes of FLGPR data and learned features are compared with several previously proposed static features. The
results suggest that learned features perform comparably, or better, than existing static features. Similar to other feature
learning results, the features consist of edges or texture primitives, revealing which structures in the data are most useful
for discrimination.
KEYWORDS: Land mines, General packet radio service, Data modeling, Detection and tracking algorithms, Target detection, Data analysis, Process modeling, Ground penetrating radar, Remote sensing, Sensors
Buried threat detection algorithms in Ground Penetrating Radar (GPR) measurements often utilize a statistical classifier to model target responses. There are many different target types with distinct responses and all are buried in a wide range of conditions that distort the target signature. Robust performance of this classifier requires it to learn the distinct responses of target types while accounting for the variability due to the physics of the emplacement. In this work, a method to reduce certain sources of excess variation is presented that enables a linear classifier to learn distinct templates for each target type’s response despite the operational variability. The different target subpopulations are represented by a Gaussian Mixture Model (GMM). Training the GMM requires jointly extracting the patches around target responses as well as learning the statistical parameters as neither are known a priori. The GMM parameters and the choice of patches are determined by variational Bayesian methods. The proposed method allows for patches to be extracted from a larger data-block that only contain the target response. The patches extracted from this method improve the ROC for distinguishing targets from background clutter compared to the patches extracted using other patch extraction methods aiming to reduce the operational variability.
Forward-looking ground penetrating radar (FLGPR) is a remote sensing modality that has recently been investigated for buried threat detection. FLGPR offers greater standoff than other downward-looking modalities such as electromagnetic induction and downward-looking GPR, but it suffers from high false alarm rates due to surface and ground clutter. A stepped frequency FLGPR system consists of multiple radars with varying polarizations and bands, each of which interacts differently with subsurface materials and therefore might potentially be able to discriminate clutter from true buried targets. However, it is unclear which combinations of bands and polarizations would be most useful for discrimination or how to fuse them. This work applies sparse structured basis pursuit, a supervised statistical model which searches for sets of bands that are collectively effective for discriminating clutter from targets. The algorithm works by trying to minimize the number of selected items in a dictionary of signals; in this case the separate bands and polarizations make up the dictionary elements. A structured basis pursuit algorithm is employed to gather groups of modes together in collections to eliminate whole polarizations or sensors. The approach is applied to a large collection of FLGPR data for data around emplaced target and non-target clutter. The results show that a sparse structure basis pursuits outperforms a conventional CFAR anomaly detector while also pruning out unnecessary bands of the FLGPR sensor.
KEYWORDS: Forward looking infrared, General packet radio service, Sensors, Systems modeling, Data modeling, Computing systems, Detection and tracking algorithms, Target detection, Cameras, Infrared sensors
Many remote sensing modalities have been developed for buried target detection, each one offering its own relative advantages over the others. As a result there has been interest in combining several modalities into a single detection platform that benefits from the advantages of each constituent sensor, without suffering from their weaknesses. Traditionally this involves collecting data continuously on all sensors and then performing data, feature, or decision level fusion. While this is effective for lowering false alarm rates, this strategy neglects the potential benefits of a more general system-level fusion architecture. Such an architecture can involve dynamically changing which modalities are in operation. For example, a large standoff modality such as a forward-looking infrared (FLIR) camera can be employed until an alarm is encountered, at which point a high performance (but short standoff) sensor, such as ground penetrating radar (GPR), is employed. Because the system is dynamically changing its rate of advance and sensors, it becomes difficult to evaluate the expected false alarm rate and advance rate. In this work, a probabilistic model is proposed that can be used to estimate these quantities based on a provided operating policy. In this model the system consists of a set of states (e.g., sensors employed) and conditions encountered (e.g., alarm locations). The predictive accuracy of the model is evaluated using a collection of collocated FLIR and GPR data and the results indicate that the model is effective at predicting the desired system metrics.
A recently validated technique for buried target detection relies on applying an acoustic stimulus signal to a patch of earth and then measuring its seismic (vibrational) response using a laser Doppler vibrometer (LDV). Target detection in this modality often relies on estimating the acoustic-to-seismic coupling ratio (A/S ratio) of the ground, which is altered by the presence of a buried target. For this study, LDV measurements were collected over patches of earth under varying environmental conditions using a known stimulus. These observations are then used to estimate the performance of several methods to discriminate between target and non-target patches. The first part of the study compares the performance of human observers against a set of established seismo-acoustic features from the literature. The simple features are based on previous studies where statistics on the Fourier transform of the acoustic-to-seismic transfer function estimate are measured. The human observers generally offered much better detection performance than any established feature. One weakness of the Fourier features is their inability to utilize local spatiotemporal target cues. To address these weaknesses, a novel automatic detection algorithm is proposed which uses a multi-scale blob detector to identify suspicious regions in time and space. These suspicious spatiotemporal locations are then clustered and assigned a decision statistic based on the confidence and number of cluster members. This method is shown to improve performance over the established Fourier statistics, resulting in performance much closer to the human observers.
KEYWORDS: General packet radio service, Forward looking infrared, Sensors, Standoff detection, Feature extraction, Detection and tracking algorithms, Data processing, Cameras, Ground penetrating radar, Land mines
Ground penetrating radar (GPR) is a popular sensing modality for buried threat detection that offers low false alarm rates (FARs), but suffers from a short detection stopping or standoff distance. This short stopping distance leaves little time for the system operator to react when a threat is detected, limiting the speed of advance. This problem arises, in part, because of the way GPR data is typically processed. GPR data is first prescreened to reduce the volume of data considered for higher level feature-processing. Although fast, prescreening introduces latency that delays the feature processing and lowers the stopping distance of the system. In this work we propose a novel sensor fusion framework where a forward looking infrared (FLIR) camera is used as a prescreener, providing suspicious locations to the GPRbased system with zero latency. The FLIR camera is another detection modality that typically yields a higher FAR than GPR while offering much larger stopping distances. This makes it well-suited in the role of a zero-latency prescreener. In this framework, GPR-based feature processing can begin without any latency, improving stopping distances. This framework was evaluated using well-known FLIR and GPR detection algorithms on a large dataset collected at a Western US test site. Experiments were conducted to investigate the tradeoff between early stopping distance and FAR. The results indicate that earlier stopping distances are achievable while maintaining effective FARs. However, because an earlier stopping distance yields less data for feature extraction, there is a general tradeoff between detection performance and stopping distance.
Forward Looking Infrared (FLIR) cameras have recently been studied as a sensing modality for use in buried threat detection systems. FLIR-based detection systems benefit from larger standoff distances and faster rates of advance than other sensing modalities, but they also present significant signal processing challenges. FLIR imagery typically yields multiple looks at each surface area, each of which is obtained from a different relative camera pose and position. This multi-look imagery can be exploited for improved performance, however open questions remain as to the best ways to process and fuse such data. Further, the utility of each look in the multi-look imagery is also unclear: How many looks are needed, from what poses, etc? In this work we propose a general framework for processing FLIR imagery wherein FLIR imagery is partitioned according to the particular relative camera pose from which it was collected. Each partition is then projected into a common spatial coordinate system resulting in several distinct images of the surface area. Buried threat detection algorithms can then be applied to each of these resulting images independently, or in aggregate. The proposed framework is evaluated using several detection algorithms on an FLIR dataset collected at a Western US test site and the results indicate that the framework offers significant improvement over detection in the original FLIR imagery. Further experiments using this framework suggest that multiple looks by the FLIR camera can be used to improve detection performance.
Roadside explosive threats continue to pose a significant risk to soldiers and civilians in conflict areas around the world.
These objects are easy to manufacture and procure, but due to their ad hoc nature, they are difficult to reliably detect
using standard sensing technologies. Although large roadside explosive hazards may be difficult to conceal in rural
environments, urban settings provide a much more complicated background where seemingly innocuous objects (e.g.,
piles of trash, roadside debris) may be used to obscure threats. Since direct detection of all innocuous objects would flag
too many objects to be of use, techniques must be employed to reduce the number of alarms generated and highlight only
a limited subset of possibly threatening regions for the user. In this work, change detection techniques are used to
reduce false alarm rates and increase detection capabilities for possible threat identification in urban environments. The
proposed model leverages data from multiple video streams collected over the same regions by first applying video
aligning and then using various distance metrics to detect changes based on image keypoints in the video streams. Data
collected at an urban warfare simulation range at an Eastern US test site was used to evaluate the proposed approach, and
significant reductions in false alarm rates compared to simpler techniques are illustrated.
Many effective buried threat detection systems rely on close proximity and near vertical deployment over subsurface
objects before reasonable performance can be obtained. A forward-looking sensor configuration, where
an object can be detected from much greater distances, allows for safer detection of buried explosive threats,
and increased rates of advance. Forward-looking configurations also provide an additional advantage of yielding
multiple perspectives and looks at each subsurface area, and data from these multiple pose angles can be potentially
exploited for improved detection. This work investigates several aspects of detection algorithms that can
be applied to forward-looking imagery. Previous forward-looking detection algorithms have employed several
anomaly detection algorithms, such as the RX algorithm. In this work the performance of the RX algorithm
is compared to a scale-space approach based on Laplcaian of Gaussian filtering. This work also investigates
methods to combine the detection output from successive frames to aid detection performance. This is done by
exploiting the spatial colocation of detection alarms after they are mapped from image coordinates into world
coordinates. The performance of the resulting algorithms are measured on data from a forward-looking vehicle
mounted optical sensor system collected over several lanes at a western U.S. test facility. Results indicate that
exploiting the spatial colocation of detections made in successive frames can yield improved performance.
In case-based computer-aided decision systems (CB-CAD) a query case is compared to known examples
stored in the systems case base (also called a reference library). These systems offer competitive
classification performance and are easy to expand. However, they also require efficient management of
the case base. As CB-CAD systems are becoming more popular, the problem of case base optimization
has recently attracted interest among CAD researchers. In this paper we present preliminary results of
a study comparing several case base reduction techniques. We implemented six techniques previously
proposed in machine learning literature and applied it to the classification problem of distinguishing
masses and normal tissue in mammographic regions of interest. The results show that the random
mutation hill climbing technique offers a drastic reduction of the number of case base examples while
providing a significant improvement in classification performance. Random selection allowed for reduction
of the case base to 30% without notable decrease in performance. The remaining techniques (i.e.,
condensed nearest neighbor, reduced nearest neighbor, edited nearest neighbor, and All k-NN) resulted
in moderate reduction (to 50-70% of the original size) at the cost of decrease in CB-CAD performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.