This PDF file contains the front matter associated with SPIE Proceedings Volume 8402, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

Hierarchy may be difficult, if not impossible, to avoid either in natural or artificially engineered systems. Nevertheless, hierarchical decomposition can be considered inconvenient in a number of ways that interfere with the construction of adaptive systems capable of full exploitation of unexpected opportunities in an operational environment. This paper will briefly review some of those inconveniences and suggest principles by which to circumvent them while maintaining the obviously beneficial, and perhaps inevitable, paradigm of hierarchical decomposition of engineering designs. The paper will illustrate the introduced principles by appeal to an example auto-adaptive flight controller for an insect-scale flapping-wing micro air vehicle. The paper will conclude with discussion of possible future applications and methodological extensions.

In both military and civilian applications, abundant data from diverse sources captured on airborne platforms are often available for a region attracting interest. Since the data often includes motion imagery streams collected from multiple platforms flying at different altitudes, with sensors of different field of views (FOVs), resolutions, frame rates and spectral bands, it is imperative that a cohesive site model encompassing all the information can be quickly built and presented to the analysts. In this paper, we propose to develop an Uncertainty Preserving Patch-based Online Modeling System (UPPOMS) leading towards the automatic creation and updating of a cohesive, geo-registered, uncertaintypreserving, efficient 3D site terrain model from passive imagery with varying field-of-views and phenomenologies. The proposed UPPOMS has the following technical thrusts that differentiate our approach from others: (1) An uncertaintypreserved, patch-based 3D model is generated, which enables the integration of images captured with a mixture of NFOV and WFOV and/or visible and infrared motion imagery sensors. (2) Patch-based stereo matching and multi-view 3D integration are utilized, which are suitable for scenes with many low texture regions, particularly in mid-wave infrared images. (3) In contrast to the conventional volumetric algorithms, whose computational and storage costs grow exponentially with the amount of input data and the scale of the scene, the proposed UPPOMS system employs an online algorithmic pipeline, and scales well to large amount of input data. Experimental results and discussions of future work will be provided.

Image registration in wide area motion imagery (WAMI) is a critical problem that is required for target tracking, image fusion, and situation awareness. The high resolution, extremely low frame rate, and large camera motion in such videos; however, introduces challenging constraints that distinguish the task from traditional image registration from such sesnors as full motion video (FMV). In this study, we propose to use the feature-based approach for the registration of wide area surveillance imagery. Specifically, we extract Speeded Up Robust Feature (SURF) feature points for each frame. After that, a kd-tree algorithm is adopted to match the feature points of each frame to the reference frame. Then, we use the RANdom SAmple Consensus (RANSAC) algorithm to refine the matching results. Finally, the refined matching point pairs are used to estimate the transformation between frames. The experiments are conducted on the Columbus Large Image Format (CLIF) dataset. The experimental results show that the proposed approach is very efficient for the wide area motion imagery registration.

Automated target tracking with wide area motion imagery (WAMI) presents significant challenges due to the low resolution, low framerate data provided by the sensing platform. This paper discusses many of these challenges with a focus on the use of features to aid the tracking process. Results illustrate the potential benefits obtained when combining target kinematic and feature data, but also demonstrate the difficulties encountered when tracking low contrast targets, targets that have appearance models similar to their background and under conditions where traffic density is relatively high. Other difficulties include target occlusion and move-stop-move events, which are mitigated with a new composite detection method that seamlessly integrates feature and kinematic data. A real WAMI dataset was used in this study, and specific vignettes will be discussed. A single target tracker is implemented to demonstrate the concepts and provide results.

In this paper, we discuss algorithmic approaches for exploiting wide-area persistent EO/IR motion imagery for multisensor geo-registration and automated information extraction, including moving target detection. We first present enabling capabilities, including sensor auto-calibration and automated high-resolution 3D reconstruction using passive 2D motion imagery. We then present algorithmic approaches for 3D-based geo-registration, and demonstrate and quantify performance achieved using public release data from AFRL's Columbus Large Image Format (CLIF) 2006 data collection and the Ohio Geographically Referenced Information Program (OGRIP). Finally, we discuss algorithmic approaches for 3D-based moving target detection with near-optimal parallax mitigation, and demonstrate automated detection of dismount and vehicle targets in coarse-resolution CLIF 2006 imagery.

We describe a comprehensive system for learning to identify suspicious vehicle tracks from wide-area motion (WAMI) video. First, since the road network for the scene of interest is assumed unknown, agglomerative hierarchical clustering is applied to all spatial vehicle measurements, resulting in spatial cells that largely capture individual road segments. Next, for each track, both at the cell (speed, acceleration, azimuth) and track (range, total distance, duration) levels, extreme value feature statistics are both computed and aggregated, to form summary (p-value based) anomaly statistics for each track. Here, to fairly evaluate tracks that travel across different numbers of spatial cells, for each cell-level feature type, a single (most extreme) statistic is chosen, over all cells traveled. Finally, a novel active learning paradigm, applied to a (logistic regression) track classifier, is invoked to learn to distinguish suspicious from merely anomalous tracks, starting from anomaly-ranked track prioritization, with ground-truth labeling by a human operator. This system has been applied to WAMI video data (ARGUS), with the tracks automatically extracted by a system developed in-house at Toyon Research Corporation. Our system gives promising preliminary results in highly ranking as suspicious aerial vehicles, dismounts, and traffic violators, and in learning which features are most indicative of suspicious tracks.

Object recognition is an important problem that has many applications that are of interest to the United States Air Force (USAF). Recently the USAF released its update to Technology Horizons, a report that is designed to guide the science and technology direction of the Air Force. Technology Horizons specifically calls out for the need to use autonomous systems in essentially all aspects of Air Force operations [1]. Object recognition is a key enabler to autonomous exploitation of intelligence, surveillance, and reconnaissance (ISR) data which might make the automatic searching of millions of hours of video practical. In particular this paper focuses on vehicle recognition with Lowe's Scale-invariant feature transform (SIFT) using a model database that was generated with semi-synthetic data. To create the model database we used a desktop laser scanner to create a high resolution 3D facet model. Then the 3D facet model was imported into LuxRender, a physics accurate ray tracing tool, and several views were rendered to create a model database. SIFT was selected because the algorithm is invariant to scale, noise, and illumination making it possible to create a model database of only a hundred original viewing locations which keeps the size of the model database reasonable.

Remotely Piloted Aircraft (RPA) are designed to operate in many of the same areas as manned aircraft; however, the limited instantaneous field of regard (FOR) that RPA pilots have limits their ability to react quickly to nearby objects. This increases the danger of mid-air collisions and limits the ability of RPA's to operate in environments such as terminals or other high-traffic environments. We present an approach based on insect vision that increases awareness while keeping size, weight, and power consumption at a minimum. Insect eyes are not designed to gather the same level of information that human eyes do. We present a novel Data Model and dynamically updated look-up-table approach to interpret non-imaging direction sensing only detectors observing a higher resolution video image of the aerial field of regard. Our technique is a composite hybrid method combining a small cluster of low resolution cameras multiplexed into a single composite air picture which is re-imaged by an insect eye to provide real-time scene understanding and collision avoidance cues. We provide smart camera application examples from parachute deployment testing and micro unmanned aerial vehicle (UAV) full motion video (FMV).

This paper was presented at the SPIE conference indicated above and has been withdrawn from publication at the request of the authors.

With the vast amount of network information available on activities of people (i.e. motions, transportation routes, and site visits) there is a need to explore the salient properties of data that detect and discriminate the behavior of individuals. Recent machine learning approaches include methods of data mining, statistical analysis, clustering, and estimation that support activity-based intelligence. We seek to explore contemporary methods in activity analysis using machine learning techniques that discover and characterize behaviors that enable grouping, anomaly detection, and adversarial intent prediction. To evaluate these methods, we describe the mathematics and potential information theory metrics to characterize behavior. A scenario is presented to demonstrate the concept and metrics that could be useful for layered sensing behavior pattern learning and analysis. We leverage work on group tracking, learning and clustering approaches; as well as utilize information theoretical metrics for classification, behavioral and event pattern recognition, and activity and entity analysis. The performance evaluation of activity analysis supports high-level information fusion of user alerts, data queries and sensor management for data extraction, relations discovery, and situation analysis of existing data.

Understanding the structure and dynamics of networks are of vital importance to winning the global war on terror. To fully comprehend the network environment, analysts must be able to investigate interconnected relationships of many diverse network types simultaneously as they evolve both spatially and temporally. To remove the burden from the analyst of making mental correlations of observations and conclusions from multiple domains, we introduce the Dynamic Graph Analytic Framework (DYGRAF). DYGRAF provides the infrastructure which facilitates a layered multi-modal network analysis (LMMNA) approach that enables analysts to assemble previously disconnected, yet related, networks in a common battle space picture. In doing so, DYGRAF provides the analyst with timely situation awareness, understanding and anticipation of threats, and support for effective decision-making in diverse environments.

In the author's previous work, a cross-layer protocol approach to wireless sensor network (WSN) intrusion detection an identification is created with multiple bio-inspired/evolutionary computational methods applied to the functions of the protocol layers, a single method to each layer, to improve the intrusion-detection performance of the protocol over that of one method applied to only a single layer's functions. The WSN cross-layer protocol design embeds GAs, anti-phase synchronization, ACO, and a trust model based on quantized data reputation at the physical, MAC, network, and application layer, respectively. The construct neglects to assess the net effect of the combined bioinspired methods on the quality-of-service (QoS) performance for "normal" data streams, that is, streams without intrusions. Analytic expressions of throughput, delay, and jitter, coupled with simulation results for WSNs free of intrusion attacks, are the basis for sensitivity analyses of QoS metrics for normal traffic to the bio-inspired methods.

With the advent of new technology in wide-area motion imagery (WAMI) and full-motion video (FMV), there is a capability to exploit the imagery in conjunction with other information sources for improving confidence in detection, tracking, and identification (DTI) of dismounts. Image exploitation, along with other radar and intelligence information can aid decision support and situation awareness. Many advantages and limitations exist in dismount tracking analysis using WAMI/FMV; however, through layered management of sensing resources, there are future capabilities to explore that would increase dismount DTI accuracy, confidence, and timeliness. A layered sensing approach enables commandlevel strategic, operational, and tactical analysis of dismounts to combine multiple sensors and databases, to validate DTI information, as well as to enhance reporting results. In this paper, we discuss WAMI/FMV, compile a list of issues and challenges of exploiting the data for WAMI, and provide examples from recently reported results. Our aim is to provide a discussion to ensure that nominated combatants are detected, the sensed information is validated across multiple perspectives, the reported confidence values achieve positive combatant versus non- combatant detection, and the related situational awareness attributes including behavior analysis, spatial-temporal relations, and cueing are provided in a timely and reliable manner to stakeholders.

Crowd motion analysis covers the detection, tracking, recognition, and behavior interpretation of target group from persistent surveillance video data. This project is dedicated to investigating a crowd motion analysis system based on a heat-transfer-analog model (denoted as CMA-HT for simplicity), and a generic modeling and simulation framework describing crowd motion behavior. CMA-HT is formulated by coupling the statistical analysis of crowd's historical behavior at a given location, geographic information system, and crowd motion dynamics. The mathematical derivation of the CMA-HT model and the innovative methods involved in the framework's implementation will be discussed in detail. Using the sample video data collected by Central Florida University as benchmark, CMA-HT is employed to measure and identify anomalous personnel or group responses in the video.

A growing body of discoveries in molecular signatures has revealed that volatile organic compounds (VOCs), the small molecules associated with an individual's odor and breath, can be monitored to reveal the identity and presence of a unique individual, as well their overall physiological status. Given the analysis requirements for differential VOC profiling via gas chromatography/mass spectrometry, our group has developed a novel informatics platform, Metabolite Differentiation and Discovery Lab (MeDDL). In its current version, MeDDL is a comprehensive tool for time-series spectral registration and alignment, visualization, comparative analysis, and machine learning to facilitate the efficient analysis of multiple, large-scale biomarker discovery studies. The MeDDL toolset can therefore identify a large differential subset of registered peaks, where their corresponding intensities can be used as features for classification. This initial screening of peaks yields results sets that are typically too large for incorporation into a portable, electronic nose based system in addition to including VOCs that are not amenable to classification; consequently, it is also important to identify an optimal subset of these peaks to increase classification accuracy and to decrease the cost of the final system. MeDDL's learning tools include a classifier similar to a K-nearest neighbor classifier used in conjunction with a genetic algorithm (GA) that simultaneously optimizes the classifier and subset of features. The GA uses ROC curves to produce classifiers having maximal area under their ROC curve. Experimental results on over a dozen recognition problems show many examples of classifiers and feature sets that produce perfect ROC curves.

In this paper we extend a previous exploration of histogram features extracted from 3D point cloud images of human subjects for gender discrimination. Feature extraction used a collection of concentric cylinders to define volumes for counting 3D points. The histogram features are characterized by a rotational axis and a selected set of volumes derived from the concentric cylinders. The point cloud images are drawn from the CAESAR anthropometric database provided by the Air Force Research Laboratory (AFRL) Human Effectiveness Directorate and SAE International. This database contains approximately 4400 high resolution LIDAR whole body scans of carefully posed human subjects. Success from our previous investigation was based on extracting features from full body coverage which required integration of multiple camera images. With the full body coverage, the central vertical body axis and orientation are readily obtainable; however, this is not the case with a one camera view providing less than one half body coverage. Assuming that the subjects are upright, we need to determine or estimate the position of the vertical axis and the orientation of the body about this axis relative to the camera. In past experiments the vertical axis was located through the center of mass of torso points projected on the ground plane and the body orientation derived using principle component analysis. In a natural extension of our previous work to partial body views, the absence of rotational invariance about the cylindrical axis greatly increases the difficulty for gender classification. Even the problem of estimating the axis is no longer simple. We describe some simple feasibility experiments that use partial image histograms. Here, the cylindrical axis is assumed to be known. We also discuss experiments with full body images that explore the sensitivity of classification accuracy relative to displacements of the cylindrical axis. Our initial results provide the basis for further investigation of more complex partial body viewing problems and new methods for estimating the two position coordinates for the axis location and the unknown body orientation angle.

The Civilian American and European Surface Anthropometry Resource (CAESAR) database containing over 40 anthropometric measurements on over 4000 humans has been extensively explored for pattern recognition and classification purposes using the raw, original data [1-4]. However, some of the anthropometric variables would be impossible to collect in an uncontrolled environment. Here, we explore the use of dimensionality reduction methods in concert with a variety of classification algorithms for gender classification using only those variables that are readily observable in an uncontrolled environment. Several dimensionality reduction techniques are employed to learn the underlining structure of the data. These techniques include linear projections such as the classical Principal Components Analysis (PCA) and non-linear (manifold learning) techniques, such as Diffusion Maps and the Isomap technique. This paper briefly describes all three techniques, and compares three different classifiers, Naïve Bayes, Adaboost, and Support Vector Machines (SVM), for gender classification in conjunction with each of these three dimensionality reduction approaches.

Entity resolution is an important area of research with a wide range of applications. In this paper we present a framework for developing a dynamic entity profile that is constructs as matching entity records are discovered. The proposed framework utilizes a fuzzy rule base that can match entities with a given error rate. A genetic algorithm is used to optimize an initial population of random fuzzy rule bases using a set of labeled training data. This approach demonstrated an F-score performance of 84% on a held out test set. The profiles that were linked demonstrated a configurable fitness measure to emphasis different search properties (precision or recall). The approach used for entity resolution in this framework can be extended to other applications, such as, searching for similar video files. Spatial and temporal attributes can be extracted from the video and an optimal fuzzy rule base can be evolved.

In recent years, position based services has increase. Thus, recent developments in communications and RF technology have enabled system concept formulations and designs for low-cost radar systems using state-of-the-art software radio modules. This research is done to investigate a novel multi-platform RF emitter localization technique denoted as Position-Adaptive RF Direction Finding (PADF). The formulation is based on the investigation of iterative path-loss (i.e., Path Loss Exponent, or PLE) metrics estimates that are measured across multiple platforms in order to autonomously adapt (i.e. self-adjust) of the location of each distributed/cooperative platform. Experiments conducted at the Air-Force Research laboratory (AFRL) indicate that this position-adaptive approach exhibits potential for accurate emitter localization in challenging embedded multipath environments such as in urban environments. The focus of this paper is on the robustness of the distributed approach to RF-based location tracking. In order to localize the transmitter, we use the Received Signal Strength Indicator (RSSI) data to approximate distance from the transmitter to the revolving receivers. We provide an algorithm for on-line estimation of the Path Loss Exponent (PLE) that is used in modeling the distance based on Received Signal Strength (RSS) measurements. The emitter position estimation is calculated based on surrounding sensors RSS values using Least-Square Estimation (LSE). The PADF has been tested on a number of different configurations in the laboratory via the design and implementation of four IRIS wireless sensor nodes as receivers and one hidden sensor as a transmitter during the localization phase. The robustness of detecting the transmitters position is initiated by getting the RSSI data through experiments and then data manipulation in MATLAB will determine the robustness of each node and ultimately that of each configuration. The parameters that are used in the functions are the median values of RSSI and rms values. From the result it is determined which configurations possess high robustness. High values obtained from the robustness function indicate high robustness, while low values indicate lower robustness.

Inherent in the Air Force's mission of airborne intelligence, surveillance, and reconnaissance (ISR) is the need to collect data from sensors. Technology is constantly advancing and, as such, new sensors are also being constantly produced. The manufacturers of these sensors typically provide with their hardware free software for communication with their sensor. These binaries work well for mature systems as interfaces and communication protocols are already firmly established. However, most research software is, by its very nature, immature and typically unable to communicate with sensor packages "out of the box." Because of this, researcher productivity is hindered as they have to focus on hardware communication in addition to their immediate research goals. As such, the creation of a library to talk to common sensors and other hardware is needed. This paper describes the various libraries currently available and their limitations. It also documents a combined effort of the Air Force Research Lab (AFRL) and Wright State University (WSU) to create a "super library" that removes as many of the limitations of each of the individual libraries as possible.

Historically, the Air Force's research into aerial platforms for sensing systems has focused on low-, mid-, and highaltitude platforms. Though these systems are likely to comprise the majority of the Air Force's assets for the foreseeable future, they have limitations. Specifically, these platforms, their sensor packages, and their data exploitation software are unsuited for close-quarter surveillance, such as in alleys and inside of buildings. Micro-UAVs have been gaining in popularity, especially non-fixed-wing platforms such as quad-rotors. These platforms are much more appropriate for confined spaces. However, the types of video exploitation techniques that can effectively be used are different from the typical nadir-looking aerial platform. This paper discusses the creation of a framework for testing existing and new video exploitation algorithms, as well as describes a sample micro-UAV-based tracker.