Big Data and Deep Analytics Applied to the Common Tactical Air
Picture (CTAP) and Combat Identification (CID)
Ying Zhao, Tony Kendall and Bonnie Johnson
Naval Postgraduate School, Monterey, CA 93943, U.S.A.
Keywords: Big Data, Deep Analytics, Common Tactical Air Picture, Combat Identification, Machine Vision, Object
Recognition, Pattern Recognition, Anomaly Detection, Lexical Link Analysis, Heterogeneous Data Sources,
Unsupervised Learning.
Abstract: Accurate combat identification (CID) enables warfighters to locate and identify critical airborne objects as
friendly, hostile or neutral with high precision. The current CID processes include processing and analysing
data from a vast network of sensors, platforms, and decision makers. CID plays an important role in
generating the Common Tactical Air Picture (CTAP) which provides situational awareness to air warfare
decision-makers. The Big “CID” Data and complexity of the problem pose challenges as well as
opportunities. In this paper, we discuss CTAP and CID challenges and some Big Data and Deep Analytics
solutions to address these challenges. We present a use case using a unique deep learning method, Lexical
Link Analysis (LLA), which is able to associate heterogeneous data sources for object recognition and
anomaly detection, both of which are critical for CTAP and CID applications.
1 INTRODUCTION
An accurate, relevant and timely CID capability
enables warfighters to locate and identify critical
airborne objects as friendly, hostile or neutral with
high precision. The objective of the CTAP is to
provide tactical situational awareness to the
decision-makers; and thereby provide critical
information to support the engagement events and
courses of action that protect Navy and Joint assets.
An effective CID and CTAP capability supports the
optimal use of long-range weapons, aids in fratricide
reduction, and ultimately reduces or minimizes
friendly forces’ exposure to enemy fire. The CID
process is an essential part of generating a CTAP.
Traditionally, CID decisions are derived from
data from intelligence, surveillance, and
reconnaissance (ISR) sensors. This research group
has noted that the size and heterogeneity of the data
from these sensors creates a Big Data environment.
The current tactical information systems cannot
meet the timelines required for CID in complex
threat environments. Nor can they process and
analyze additional types of data that may support
CID, such as information from the Internet, social
media, and commercial airline information. We are
studying new methods such as Big Data and Deep
Analytics that show promise for handling and
analyzing the rising tide of sensor and non-sensor
data in a timely manner.
The Aegis combat system, CEC, and Link 16 are
critical systems supporting CID for sharing data
among distributed platforms, correlating and fusing
data, and displaying airborne object tracks.
Additionally, the current CID processes include the
use of Naval CTAP components and combinations
of:
Platforms: destroyers, cruisers, carriers, F/A-
18s, E-2C/D, LHD/LHAs and Amphibious
Assault Ships.
Sensors: radar, Forward Looking Infrared
(FLIR), Identification Friend or
Foe (IFF),
Precision Participation Location Identifier (PPLI),
and National Technical Means (NTM)
Networks: Cooperative Engagement Capability
(CEC), Link-16 Global Command and
Control
System (GCCS), and Global Information Grid
(GIG)
Decision makers: Air and Missile Defense
Commander (AMDC), Air Warfare (AW)
Offi
cer, Tactical Action Officer (TAO) and Air
Defense Officer (ADO)
Zhao, Y., Kendall, T. and Johnson, B.
Big Data and Deep Analytics Applied to the Common Tactical Air Picture (CTAP) and Combat Identification (CID).
DOI: 10.5220/0006086904430449
In Proceedings of the 8th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2016) - Volume 1: KDIR, pages 443-449
ISBN: 978-989-758-203-5
Copyright
c
2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
443
The challenges for CTAP and CID include:
An extremely short dwell time for fusion,
decision making, and targeting.
Uncertain and/or missing data outside sensor
ranges (e.g., radar). For example, track pictures
are uncertain with track conflicts, multiple
objects per track or multiple tracks per object.
Manual decision-making. For example, complex
threat environments can create situations in
which decision-makers can be overwhelmed by
large amounts of data, uncertain track pictures,
and complicated doctrine.
Hard-to-detect anomalies and a lack of predictive
analytic capabilities.
Manual methods for incorporating electronic
warfare (EW), electronic intelligence (ELINT),
and non-cooperative sensor measurements and
signature databases, into the CID process.
The contribution of this paper is to position
various Big Data and Deep Analytics in the context
of Big “CTAP and CID” Data. We also show a
unique Deep Learning method, i.e., Lexical Link
Analysis (LLA), which uses a bi-gram model to link
any two entities across multiple contexts and
associate heterogeneous data sources for object
recognition and anomaly detection.
2 BIG DATA
2.1 Big Data Problem
Today, Big Data is omnipresent. Big Data science
intervenes with traditional data sciences. We are
compelled to ask - What is new? Here, we examine
some aspects of the problem:
Big rise in data: Data creation is remarkable for
its volume, velocity, and variety. “Volume”
considers the rise of new data creation platforms
of multimedia, social media, mobile devices, the
Internet of Things (IOT) and new sensors.
“Velocity” considers these new platforms
capturing millions of events per second and in
real-time. “Variety” considers captured data are
also unstructured text, images, audios, videos,
geospatial data, and 3D data.
Big rise in needs: It is critical for business to
transform data into smart data, or actionable
knowledge.
Big rise in analytics: Traditional data sciences
including statistics, numerical analysis, machine
learning, data mining, business intelligence, and
artificial intelligence have evolved into Big
Data analytics or Deep Analytics. These
technologies can be overwhelmingly complex,
requiring diversified and extensive expertise.
2.2 Tools and Challenges
Big Data requires massively parallel software on
thousands of servers. The current technologies are
dominated by systems that provide 1) data
collection, ingestion, integration and safe storage; 2)
parallel/distributed processing; and 3) Deep
Analytics.
As part of the open-sourced Apache Hadoop
ecosystem, Hadoop Distributed File System (HDFS)
provides distributed and fault-tolerant data storage.
Beehive and Pig are "SQL-like" tools for
conventional database queries on a HDFS. NoSQL
systems include document and graph databases in a
“cloud” such as Amazon and Cloudera. NoSQL
databases are increasingly used because of
simplicity of design, horizontal scaling, and finer
control over availability.
Operational systems for messaging, banking,
advertising and mobile devices can utilize Apache
Storm to handle day-to-day transactions in real-time,
or with no- or low-latency of response.
Map/Reduce is an analytic programming
paradigm for Big Data. It consists of two tasks: 1)
the "Map" task, where an input dataset is converted
into key/value pairs; and 2) the "Reduce" task,
where outputs of the "Map" task are combined to a
reduced key-value pairs. Apache Spark (Spark,
2016) is replacing Map/Reduce for its speed and in-
memory computation.
As the data size gets bigger, the statistical
significance for an analysis is often guaranteed due
purely to the data size. This positive impact of the
data size can be a great advantage. However, other
challenges rise. For example, traditional data
sciences used in small- or moderate-sized analysis
typically require tight coupling of the computations
of the “Map” and “Reduce” steps. Such an algorithm
often executes in a single machine or job and reads
all the data at once. How can these algorithms be
modified so they can be executed in parallel in
thousands of clusters?
2.3 Big CTAP and CID Data
Data sources for Department of Defense (DoD)
applications including disparate, multi-sourced real-
time sensors are of extremely high rates and large
volumes. In DoD collaboration environments, the
needs for information sharing and agility as well as
KDIR 2016 - 8th International Conference on Knowledge Discovery and Information Retrieval
444
strict security across all domains make the matter
more complex. While commercial applications such
as massive marketing may require identifying
information with popular and repeatable patterns,
emerging and anomalous information are more
useful for DoD applications (e.g., intelligence
analysis and resource management). Deep learning
regarding pattern recognition, anomaly detection,
and data fusion can be even more useful. The US
Navy has now begun to take initiatives to move Big
Bata into the battlefield (NBD, 2014).
The data used for CID come from a combination
of massive cooperative and non-cooperative sensors,
organic sensors and non-sensor information. In
reality, each sensor collects certain attributes. The
Big CID Data need to be fused over time and space
since they are collected in a distributed fashion as
shown in Figure 1.
Figure 1: A holistic view of Big CTAP and CID Data.
3 DEEP ANALYTICS
3.1 Commercial Trends
It is critical to turn Big Data into smart data. One
important trend is Deep Analytics including analytic
algorithms that can be run in parallel and distributed
fashion.
Predictive analytics turns Big Data into smart
data, for example, accurately forecasting high-value
targets. The topic has been thoroughly studied in
traditional supervised learning. Some algorithms are
implemented using Big Data and Deep Learning
requirements such as Map/Reduce paradigm,
Mahout (2016) and Spark, (2016).
Social network analysis and graph search require
graph analyses leveraging massively parallel
processors. Graph algorithms can process petabytes
of data and are considered as the core drivers of Big
Data. Spark, Titan and Neo4j are used for Big
Graph.
3.2 Deep Learning
Deep Learning models, in a nutshell, are much
larger machine learning models with many more
parameters that are specifically designed to handle
Big Data. Deep Learning models including Deep
supervised machine learning models, e.g., convo-
lutional neural networks (CNN, 2016) with much
deeper hidden layers; Deep reinforcement learning
models; and Deep unsupervised machine learning
models for recognizing objects and patterns of
interest. Sparse coding (Olshausen and Field, 1996)
and self-taught learning (Le, Ranzato, Monga,
Devin, Chen, Corrado, Dean, and Ng, 2012) make
Deep unsupervised learning possible. The self-
taught learning is also a deep unsupervised learning
model that approximates the input for unlabelled
objects as a succinct, higher-level feature representa-
tion of sparse linear combination of the bases. It uses
the Expectation and Maximization (EM) method to
iteratively learn coefficients and bases (LeCun,
Bottou, Bengio, and Haffner, 1998). Deep Learning
models links machine vision and text analysis smar-
tly. For example, Latent Dirichlet Analysis (LDA,
Blei, Ng and Jordan, 2003) is a sparse coding where
a bag of words used as the sparsely coded features
for text (Raina, Battle, Lee, Packer and Ng, 2007).
Our methods Lexical Link Analysis (LLA, Zhao,
Gallup and Mackinnon, 2011, 2015), System-Self-
Awareness (SSA, Zhao and Zhou, 2016), and Colla-
borative Learning Agents (CLA, Zhou, Zhao and
Kotak, 2009) can be viewed as Deep models, in the
sense similar to the LDA method as a Deep Learning
method (Raina, Battle, Lee, Packer and Ng, 2007).
4 DEEP ANALYTICS FOR CID
4.1 The CTAP Cloud Concept
We first explored how Big Data and Deep Analytics
could address the challenges of CID. We developed a
CTAP Cloud Concept as shown in Figure 2.
Conceptually, it can be physically associated with
a Big Data cloud implementation such as the Naval
Tactical Cloud (NTC). It could store traditional
CTAP and CID data sources as well as the
additional non-traditional data sources, such as
temporal, spatial and organic sensor data that are
collected but not currently used (e.g. Aegis residual
data), open sources flight schedules, advanced
Big Data and Deep Analytics Applied to the Common Tactical Air Picture (CTAP) and Combat Identification (CID)
445
(EW/ELINT) signature data sources, and intelligence
data. These new data sources could be fused and
analyzed in parallel using Deep Analytics in a CTAP
Cloud. The resulting knowledge repository, i.e.,
smart data, could be searched, matched, and cross-
validated with real-time new data streams. For
example, the cloud could send or push the smart data
(e.g. early warnings or alerts) to various platforms
within a battlespace. A platform with partial or
uncertain sensor/track data could send a real-time
query to the cloud to find a higher certainty match.
The smart data push and pull would have a relatively
small data size and therefore not strain current
networks for transmission between platforms.
Figure 2: The CTAP Cloud Concept.
The CID/CTAP application domain is an
extremely complex and amazingly interesting field
in terms of the roles that many Big Data and Deep
Models can play. We investigated Big Data and Deep
Analytics to address CTAP and CID challenges
including the following areas:
Machine vision and Deep Learning models: These
algorithms have the potential to improve object
recognition, classification accuracy and probability
of correctly identifying air objects
by associating,
correlating, and fusing heterogeneous data
sources that do not share data models. This
process is demonstrated with unclassified tactical
data samples of infrared (IR) and
Electro-optical
(EO) images in this paper (Section 4.1).
Pattern recognition, anomaly detection and
unsupervised learning models: We developed and
selected pattern recognition and anomaly
detection algorithms that could be used for
identifying intent, air picture event anomalies or
launch predictions.
Optimization, decision making and deep
reinforcement learning models: We investigated
Big Data optimization, decision making and
reinforcement learning models such as Q-learning
in Soar (2016) and DeepMind (2016) that can be
used for CTAP and CID. The models could not
only automate many current manual CTAP and
CID processes but also have the potential to enhan-
ce future CTAP capabilities such as uncooperative
game theory and total battle management.
Fast, parallel and distributed computing models:
Commercial tools for Big Data may not satisfy
CTAP and CID which requires fast, parallel and
distributed computing. Tools such as associative
arrays (Kepner, Chaidez, Gadepally and Jansen,
2014), BigDAWG polystore (2016) and
GraphBLAS (2016) may have the potential to
address the requirements.
4.2 Machine Vision and LLA
LLA is an unsupervised deep learning method,
implemented in parallel and distributed fashion. By
using LLA, a complex system can be expressed in a
list of attributes or features with specific
vocabularies or lexicon terms to describe its
characteristics and surrounding environment. LLA
uses
bi-gram word pairs, compared to LDA, are
potentially more meaningful and sparse coded
features. Specifically, LLA is a form of text analysis.
For example, word pairs or bi-grams as lexical terms
and
features can be extracted and learned from a
document repository. For a text document, words are
represented as nodes and word pairs as the links
between nodes. Figure 3 shows an example of such
a word network, for example, “cash dividend”,
“dividend report”, and “market influence” are
examples of bi-gram word pairs from a financial
news data sample. LLA is related to Latent Semantic
Analysis (LSA, Dumais, Furnas, Landauer and
Deerwester, 1988), Probabilistic Latent Semantic
Analysis (PLSA, Hofmann, 1999), WordNet (Miller,
1995), Automap (CASOS, 2009), and LDA (Blei,
Ng and Jordan, 2003). LDA uses a bag of single
words (e.g., associations are computed at the word
level) to extract concepts and
topics. LLA uses bi-
gram word pair. LLA was previously used in many
examples for understanding DoD data (Zhao,
McKinnon and Gallup, 2009, 2011, 2015).
The unique characteristic of LLA is that the Bi-
gram also allows it to be extended to data other than
text (e.g., numerical or categorical data). For
example, structured data from databases can be
discretized or categorized to word-like terms. For
KDIR 2016 - 8th International Conference on Knowledge Discovery and Information Retrieval
446
example, features, such as age_older_than_65 and
gender female can be generated from the age
and gender attributes.
The word pair model can further be extended to
a context-concept-cluster (CCC, Zhao and Zhou,
2014) model. A context is a word or attribute that are
shared by multiple data sources. A context can be a
location, a time point or an object that are shared
across data sources. Using this generalization, a bi-
gram or word pair model can used to link any two
entities across multiple contexts. This is the key
point for LLA used in the CTAP and CID analytics
to associating heterogeneous data sources (see the
use case in Section 4.3).
Figure 3: An example of a theme or topic discovered by
LLA for a text data.
4.3 Use Case
4.3.1 Data Samples
The sample data contains a large collection of
visible and IR imagery collected by the US Army
Night Vision and Electronic Sensors Directorate
(NVESD). It contains 207 GB of IR imagery and
106 GB of visible imagery along with an image
viewer, ground truth data, meteorological data,
photographs of the objects, and other documentation
to assist the user in correctly interpreting the
imagery. All imagery was taken using commercial
cameras operating in the IR
and visible bands.
The data was pre-processed using SiFT-like code
(SiFT,2016)
to generate 400 visual “words
(histograms to the centers of k-means) so LLA bi-
gram models can be
applied. Figure 4 summarizes
the processed data, consisting of 4500 total training
images with 400 features
or visual words for nine
classes of objects (target vehicles) and two different
modalities (i.e., IR and EO sensors). Therefore with
4500 total images per test, there were a total of 9000
images. Each object in each mode contained 500
images.
The baseline object recognition for this
data was given using the method of representation
learning through topic models (Flenner, 2015).
Figure 4: Images data were pre-processed to feed to LLA.
4.3.2 Associating Data Sources
Another challenge to improving CID is that traditional
ISR sensor data does not have standardized or
common data attributes; and often there are missing
attributes. For example, IR and EO sensors use
completely different features (vocabularies). We used
a generalized LLA model of bi-gram co-occurrence
of spatial locations (i.e., image patches) to link two
modalities. For example, an IR image feature (i.e.,
the concept in a CCC model) describes the same
image characteristics with an EO image feature
because these two features are frequently used in the
same image patches (i.e., contexts in the CCC model).
This learning paradigm is a generic framework to fuse
two data sources. The data sources do not share
vocabularies and some data are even missing or
uncertain. Nevertheless, they can all be fused into
one picture using this method.
4.3.3 Applying LLA
We applied LLA to the data set as follows:
Step 1: Divide data into a training data set
and a test data set: each object has 500 images
which are divided into 250 images for training and
250 images for test. Bi-gram and association
learning are performed on 250 training images.
There are 36 data sets of nine training sets and nine
test sets for the two modalities for the nine objects.
Step 2: Extract bi-gram features for each data
set in a distributed fashion. Uni-gram or
bi-gram
features for each of 36 data sets are then processed
separately.
Big Data and Deep Analytics Applied to the Common Tactical Air Picture (CTAP) and Combat Identification (CID)
447
Figure 5: Target types.
An unsupervised learning system ideally should
discover nine clusters. According to the data
description in Figure 5, some of the nine (0-8)
objects are very similar in nature. An automatic
unsupervised method is expected to see fewer
clusters of objects. Figure 5 shows that 36 data
sets are grouped into five clusters.
4.3.4 Results
We first applied the “uni-gram setting: this is
related to the “bag-of words” approach where only
the 400 features are used to distinguish the objects.
The correlation for any of the two data sets is flat
and similar, indicating a uni-gram or a bag-of-words.
This indicates that the 400 features are not good for
separating, recognizing and distinguishing these
objects.
Figure 6: LLA discovered five clusters of objects.
The second setting of LLA we used generated
both full bi-gram and association learning between
IR and EO. This is shown in Figure 6. There are five
clusters for nine classes of the objects as follows:
Cluster 1: class 0 (pick up) and class 1 (sport
utility vehicle)
Cluster 2: class 2 (infantry scout vehicle), class
3 (armored personal carrier) and class 8
(armored reconnaissance vehicle towing a D20
artillery piece)
Cluster 3: class 4 (armored personal carrier)
Cluster 4: class 5 (main battle tank) and class 6
(anti-aircraft weapon)
Cluster 5: class 7 (self-propelled howitzer)
Five clusters are consistent with the ones marked
in Figure 5. Initial results in the use case show Deep
Analytics such as LLA can automatically discover
categories of objects in a Big Image Data.
5 FUTURE WORK
Our team plans to combine and test sample CID
track data with FAA and twitter data. We will test
several behavior-based Deep Learning algorithms to
see if there are normal patterns and anomalies for the
military aircraft and commercial ones. The goal is
to see if added databases and Deep Analytics will
improve CID and the CTAP.
6 CONCLUSIONS
We identified and assessed the current CTAP and
CID Big Data problems and challenges; and
identified key Deep Analytics required to address the
challenges. Big Data and Deep Analytics were found
to have potential in improving object recognition
and classification through the utilization of more
databases, distributed computation, and data fusion.
These applications could be realized by the adoption
of our cloud architecture concept which includes
continuous monitoring in time and space; and
collecting and processing data in a cloud. Finally, the
team found that the unique LLA method is able to
associate heterogeneous data sources and perform
Deep unsupervised Learning; which implies a future
application to the CID and CTAP.
ACKNOWLEDGEMENTS
Thanks to our research sponsors, Mr. William A.
Treadway and Mr. Richard Heathcote from the
OPNAV Combat Identification Capability Organiza-
tion. Thanks to Dr. Arjuna Flenner in the U.S. Naval
Air Warfare Center, who provided insightful domain
expertise and discussion for the research. Thanks to
the Naval Postgraduate School Research Program
for funding this project.
KDIR 2016 - 8th International Conference on Knowledge Discovery and Information Retrieval
448
REFERENCES
Blei, D., Ng, A. and Jordan, M. 2003. Latent Dirichlet
allocation. Journal of Machine Learning Research,
3:993-1022. Retrieved from http://jmlr.csail.mit.edu/
papers/volume3/blei03a/blei03a.pdf.
BigDAWG, 2016. A Demonstration of the BigDAWG
polystore system. http://livinglab.mit.edu/wp-content/
uploads/2016/01/bigdawg-polystore-system.pdf.
Center for Computational Analysis of Social and
Organizational Systems (CASOS) 2009. AutoMap:
extract, analyze and represent relational data from
texts. Retrieved from http://www.casos.cs.cmu.edu.
DeepMind, 2016. https://deepmind.com/
Dumais, S. T., Furnas, G. W., Landauer, T. K. and
Deerwester, S. 1988. Using latent semantic analysis to
improve information retrieval. In Proceedings of
CHI’88: Conference on Human Factors in Computing,
281-285.
Freeman, L.C. 1979. Centrality in social networks I:
conceptual clarification. Social Networks, 1: 215-239.
Flenner, A. 2015. Representation learning through topic
models. NFCS NAWCWD, China Lake, in the 2015
National Fire Control Symposium.
GraphBLAS, 2016. Graph algorithms for basic linear
algebra subprograms (BLAS). http://istc-bigdata.org/
GraphBlas/
Hofmann, T. 1999. Probabilistic latent semantic analysis.
Proceedings of the Fifteenth Conference on Uncertain-
ty in Artificial Intelligence, Stockholm, Sweden.
Kepner J., Chaidez, J., Gadepally, V., Jansen, H. 2014.
Associative Arrays: Unified Mathematics for
Spreadsheets, Databases, Matrices, and Graphs.
http://db.csail.mit.edu/nedbday15/pdf/Paper7.pdf.
LeCun, Y., Bottou, L., Bengio, Y. and Haffner, P. 1998.
Gradient-based learning applied to document recog-
nition. Proceedings of the IEEE, 86(11):2278-2324.
Retrieved from
http://yann.lecun.com/exdb/publis/pdf/lecun- 01a.pdf.
Miller, G. A. 1995. WordNet: a lexical database for
English. Communications of the ACM, 38(11).
Mahout, 2016. http://mahout.apache.org/
NBD, 2014. Navy Big Data. http://defensesystems.com/
articles/2014/06/24/navy-onr-big-data-ecosystem.aspx.
Olshausen, B. and Field, D. 1996. Emergence of simple-
cell receptive field properties by learning a sparse code
for natural images. Nature.
Raina, R., Battle, A., Lee, H., Packer, B., and Ng, A.Y.
2007. Self-taught learning: transfer learning from
unlabeled data. In ICML.
Ranzato, M., Monga, R., Devin, M., Chen, K., Corrado,
G., Dean, J., Le, Q. V. and Ng, A.Y. 2012. Building
high-level features using large scale unsupervised
learning. Proceedings of the 29th International
Conference on Machine Learning (ICML-12).
Retrieved from http://arxiv.org/pdf/1112.6209v5.pdf.
Soar, 2016. http://soar.eecs.umich.edu/
Spark, 2016. http://spark.apache.org/
SiFT, 2016. http://www.vlfeat.org/
Zhou, C., Zhao, Y., and Kotak, C. 2009. The Collaborative
Learning agent (CLA) in Trident Warrior 08 exercise.
In Proceedings of the International Conference on
Knowledge Discovery and Information Retrieval
(KDIR), Madeira, Portugal.
Zhao, Y., Gallup, S.P. and MacKinnon, D.J. 2011. System
self-awareness and related methods for improving the
use and understanding of data within DoD. Software
Quality Professional, 13(4): 19-31. http://asq.org/
pub/sqp/
Zhao, Y., Mackinnon, D. J., Gallup, S. P. 2015. Big data
and deep learning for understanding DoD data. Journal
of Defense Software Engineering, Special Issue: Data
Mining and Metrics.
Zhao, Y., Mackinnon, D. J., Gallup, S. P. 2015. Big data
and deep learning for understanding DoD data. Journal
of Defense Software Engineering, Special Issue: Data
Mining and Metrics.
Zhao, Y., Zhou, C. 2016. System Self-Awareness Towards
Deep Learning and Discovering High-Value
Information. The 7th IEEE Annual Ubiquitous
Computing, Electronics and Mobile Communication
Conference, New York City, USA 20 - 22 October
2016.
Zhao, Y. and Zhou, C. 2014. System and method for
knowledge pattern search from networked agents. US
patent 8,903,756. https://www.google.com/patents/
US8903756.
Big Data and Deep Analytics Applied to the Common Tactical Air Picture (CTAP) and Combat Identification (CID)
449