Amelican Defense
`Preparedness Association's
`Security Technology Division
`
`'
`
`.i..
`
`•
`
`•
`
`,
`
`\
`
`10th
`Joint Annual
`Government-Industry
`Securit
`-~ Techno ogy
`
`Symposium &
`Exhibition
`",Reducing Security Vulnerabilities
`Through Innovation and Technology-(cid:173)
`The Risk Management Challenge"
`
`.. .
`
`.
`
`' I
`I
`I
`
`I
`
`I
`
`l
`I I
`
`Williarr1sburg, Virginia
`June 20-23,. 1994
`----
`
`:re
`
`~
`
`. .
`
`

`

`J (
`
`-,
`
`TABLE OF CONTENTS
`
`~ PRESENTATIONS
`
`PAGE
`
`SESSION I
`Government Security Perspectives
`
`PANEL PRESENfATION
`Domin;c J. Monetta
`Resources Alternatives Inc. - Washington, DC
`
`WE HA VE MET THE ENEMY AND fT IS US
`Maurice N. Shriber .
`System Planning Corp. - Williamsburg, VA
`
`TERRORISM IN THE NINETIES
`Robert H. KuppBrman
`Georgetown University/Centre for S&/1 - Washington, DC
`
`SESSION II
`Transportation Security
`
`DEPARTMENT OF TRANSPORTATION SECURITY INTELLIGENCE
`RADM Paul Busick
`Dept. of Transportation/USCG - Washington, DC
`
`X-RAY CARGO INSPECTION
`Dip/. Ing. Fred Hemp
`Heimann Systems GmbH - Germany
`
`DEFENSE TRANSPORTATION TRACKING SYSTEM (DITS}
`Gary Henning -
`Naval Ordnance Command -_ Indian Head, MD
`
`RAIL SHIPMENT SECURITY
`J. P. McMahon
`Police & Special Services - Jacksonville, FL
`
`TRANSPORTATION TERRORISM: NEW THREATS & VULNERABILITIES
`Edward V. Badolato
`Management Services, Inc. - Falls Church, VA
`
`SESSION Ill
`Risk Management
`
`PHYSICAL SECURfTY TECHNOLOGIES FOR WEAPONS
`COMPLEX RECONFIGURATION FACILfTES
`Calvin D. Jaeger
`Sandia National Laboratories - Albuquerque, NM
`
`1
`
`5
`
`13
`
`27
`
`33
`
`41
`
`47
`
`51
`
`55
`
`

`

`TABLE OF CONTENTS (Continuation)
`
`PRESENTATIONS
`
`PAGE
`
`SAVING MONEY & RESOURCES THROUGH TECHNOLOGY
`(SMARTT} ASTRA TEGIC PLANNING INfTIAT1VE
`Richard J. Certo
`Lawrence Livermore National Lab. - Livermore, CA
`
`APPL YING RISK MANAGEMENT AND PROGRAM PROTECTION: PLANNING
`IN THE JOINT ADVANED STRIKE TECHNOLOGY (JAST} PROGRAM
`Major Ken Newsham & Ms. Jennifer Mastroianni
`JAST Program Office - Arlington, VA
`
`SESSION IV
`Department of Defense Programs
`
`DOD SECURITY PROORAMS
`Michael Toscano
`PSEAG - Washington, DC
`
`US AIR FORCE SECURl1Y SYSTEMS ACQUISITION
`Douglas W. Dalessio
`Security Systems Product Group - Hanscom AFB, MA
`
`REDUCING SECURfTY VULNERABILJTIES THROUGH INNOVATION
`& TECHNOLOGY - THE RISK MANAGEMENT CHALLENGE
`Leopold L. Targosz, Jr
`·
`NA VCRIMINVSERV • Washington, DC
`
`i•" ..... _
`
`US ARMY AVIATION & TROOP COMMAND
`Lieutenant Colonel Bernard Wilson
`OPSEMO - Fort Be/voir, VA
`
`DNA PHYSICAL SECURITY EQUIPMENT PROGRAMS
`Lieutenant Colonel Johnnie J. Gore
`HQ DNA (NOSA} - Alexandria, VA
`
`t:
`
`i;
`~
`·· ·
`)
`~-~"- _,, . ... '"'
`
`THE AIR MOBILE GROUND SECURITY & SURVEIUANCE SYSTEM (AMGSSS)
`D.W. Murphy
`NCC&OSCRDT&E Division - San Diego, CA
`
`ARMY SECURITY STANDARDS & CRITERIA TO RESIST
`CRIMINAL & TERRORIST THREATS
`Mary Nelson Darling
`US Army Corps of Engineers - Omaha, NE
`
`WIDE AREA RAMP SURVEILLANCE SYSTEM - A LOW COST
`AL/TOMA TED THERMAL IMAGING SYSTEM
`Glenn E. Herosian
`USAF System Resource Corp. - Hanscom AFB, MA ~
`
`63
`
`73
`
`83
`
`89
`
`99
`
`107
`
`121
`
`129
`
`135
`
`143
`
`

`

`TABLE OF CONTENTS (Continuation)
`
`PRESENTATIONS
`
`DOD SECURITY EQUIPMENT TESTING: FACILITIES,
`CAPAB/Lff/ES & AVAILABILITY
`ShMey Mattingly
`USN, NCIS - Washington, DC
`
`SESSION V
`Information & Computer Security
`
`LOW COST IMAGE TRANSMISSION SYSTEM
`David Skogmo
`Sandia National Laboratories - Albuquerque, NM
`
`COMPtnER CRIME & SECURITY INCIDENT STUDY -sDN-OF-Sl..AMMER•
`Special Agent Jim Christy
`Air Force Office of Special Investigations
`
`SYSTEM SECLIRITY ENGINEERING
`Virgil L. Gibson
`CISSP, Grumman Data Systems - Linthicum, MD
`
`BACK TO BASIS: HOW MUCH COMPUTER SECURITY DO WE
`REAl..1.. Y NEED?
`Hays W. McCormick Ill & Michael J. Hoagland
`Space Applications Corporation - Vienna, VA
`
`SESSION VI
`Department of Justice
`
`PERIMETER SECURITY TRENDS IN EUROPEAN PRISONS
`Robert M. Rodger
`Police Scientific Development Branch - United Kingdom
`
`DYNAMIC HIGH PERFORMANCE ACCESS CONTROL IN
`OBJECT-ORIENTED SYSTEM
`P,eter Shaohua Deng
`Central Police University - Taiwan, ROG
`
`AUTONOMOUS SCENE MONffORING SYSTEM
`Bruce Flinchbaugh & Tom Bannan
`Texas Instruments - Dallas, TX
`
`PAGE
`
`149
`
`159
`
`165
`
`177
`
`183
`
`193
`
`199
`
`205
`
`211
`
`SESSION VII
`Arms Control & Non-Proliferation
`OPERATIONS SECURITY & ARMS CONTROL TREATY IMPLEMENTATION
`Henry H. Horton
`Dyncorp Meridian Corp. - Alexandria, VA
`
`---------------------
`
`

`

`TABLE OF CONTENTS (Continuation)
`
`PRESENTATIONS
`
`PAGE
`
`SES_SION VIII
`Security Systems Engineering
`
`PANORAMIC IMAGING PERIMETER SENSOR DESING & MODELING
`Danial A. Pritchard
`Sandia National Laboratories - Albuquerque, NM
`
`DEMYSTIFYING TECHNOLOGY TRANSFER: AVAILABLE SECURITY
`TECHNOLOGY & GUIDELINES FOR MANAGING THE PROCESS
`Neal Owens
`BA TTELLE - Columbus, OH
`
`FENCES & FENCE SENSORS IN DELAY & DETECTION
`Dr.- Mel C. Maki
`Senstar Corporation - Kanata, Ontario, Canada
`
`INTRUSION SENSOR AUTOMATED TESTING
`David R. Hayward
`Sandia National Laboratories - Albuquerque, NM
`
`ADVANTAGES OF REDEPLOYABLE & RELOCA'TABLE
`SECURITY SYSTEMS
`-
`Arthur Birch
`.
`ECSl-lnternational - Fairfield, NJ
`
`MODERNIZING YOUR SECURITY SYSTEM WITH ROBOTICS
`Celesta Decorte
`Cybermotion Inc. -Salem, VA
`
`IRIS IDENTIRCATIONIVERIRCA TION TECHNOLOGY
`G. 0 . Wi/llams, lriScan, Inc. - Mt. Laurel, NJ
`John Daughman, Cambridge University - United Kingdom
`
`HIGH CONRDENCE PERSONAL IDENTIFICATION BY.RAPID
`VIDEO ANALYSIS OF IRIS TEXTURE
`_
`John Daughman, Ph.D.
`Cambridge University & lriScan, Inc. - Mt. Laurel, NJ
`
`SECURITY SOLUTIONS AND LESSONS LEARNED FOR THE B-2
`Joa Dixon
`ESCIAVJ - Hanscom AFB, MA
`
`BIOMETRIC FACIAL RECOGNfTION USING INFRARED IMAGERY
`David C. Evans
`Protection Programs & Technologies - Alexandria, VA
`
`TACTICAL AlffOMATED SECURITY SYSTEMS (TASS)
`1 Lieutenant Dave Damrath
`Security Systems Product Group - Hanscom AFB, MA
`
`219
`
`225
`
`231
`
`239
`
`245
`
`249
`
`257
`
`280
`
`292
`
`300
`
`314
`
`

`

`TABLE OF CONTENTS (Continuation)
`
`PRESENTATIONS
`
`NEURAL NETWORK TECHNOLOGY. THE NEXT STEP FOR PERIMETER
`SECURITY INTRUSION DETECTION
`J. Somers, Dr. A. Sanders, B. Skiffington & P. Mogensen
`SWL, Inc. - Vienna, VA
`
`ADVANCED TECHNOLOGY STRENGTHENS NEW INTRUSION
`DETECTION SENSORS AGAINST INTERNAL SABOTAGE
`Jerry Outslay
`·
`.
`Sentrol, Inc. - Tualatin, OR
`
`ALTERNATE PAPERS
`
`THE ARMY INTEGRATED RISK & THREAT ANALYSIS PROCEDURES
`Curt P. Betts
`US Army Corps of Engineers - Omaha, NE
`•EYE ON THE THREA r A VIDEO REMOTE TRANSMISSION SYSTEM
`Henry F. Daidone
`SWL, Inc. - Vienna, VA
`
`A LONG LIFE TAMPER INDICATOR
`Edward F. Divers Ill - ft. Meade, MD
`
`ATTENDANCE ROSTER
`
`PAGE
`
`324
`
`332
`
`340
`
`346
`
`352
`
`358
`
`

`

`SESSION VI
`Department of Justice
`Speakers will focus on the new technologies &
`policies developing in light of the modern challenges and
`risks in law enforcement and security policy
`Session Chairperson:
`James Mahan, Session Chairperson
`Federal Bureau of Prisons - Washington, DC
`
`

`

`Autonomous Scene Monitoring System
`
`Bruce Flinchbaugh & Tom Bannon
`
`Texas Instruments
`Systems & Information Science Laboratory
`P.O. Box 655474, MS 238
`Dallas, TX 75265
`Phone: 214-995-0349
`
`Abstract
`
`Autonomous scene monitoring poses substantial requirements for extracting, storing, and
`displaying information ab~ut three-dimensional (3-D) scenes. We have developed and
`demonstrated a data base system ca.lled visual men;iory that interfaces automated video
`camera monitoring systems with ep.d-user applications requiring real-time and historical
`information about observed objects. T~is paper describes our visual memory a.nd real(cid:173)
`time camera monitoring capabilities. The visual memory prototype uses state-of-the-art
`object-oriented data base technology with spatio-temporal indexing extensions. The
`video monitoring system reports 3-D positions of people in the field of view of a CCD
`camera to visual m_emory, which dynamically maintains the information with respect to
`a map, and a user interface proyides interactive access to the data via historical ~d
`real-time queries.
`
`1 Motivation
`
`CCTV surveillance cameras provide valuable data in many security monitoring situations.
`For example, in some cases CCTV images are recorded using time lapse video recorders.
`Security system opera.tors use CCTV monitors for remote situation assessment when an
`ala.rm detector signals a potential problem. And in some systems video.__ motion detectors
`are used to automatka.lly signal alarms when changes are detected in the CCTV data.
`lnterestingly, practically all of the information available from CCTV cameras is essen(cid:173)
`tially ignored. In many time-lapse video surveillance situations, the images recorded on
`tape are never viewed or used in any way m1less a specific event occurs, such as a.n acci(cid:173)
`dent or a theft, and even then only a small portion Qf the data. may be viewed. Although
`people are generally good at assessing situations using CCTV data, it is well known that
`operator performance deteriorates significantly with fatigue. And video motion detectors
`
`205
`
`I
`I
`II
`I
`I
`
`II
`
`

`

`Visual Memory:
`
`Vision
`System
`
`User
`Interface
`
`Vision System:
`
`Visual Memory:
`
`User Interface:
`
`People Detection
`
`3-D Localization
`&Tracking
`
`Spatio-Temporal
`Indexes, Queries,
`& Object Classes
`
`Real-1ime Displays
`
`Historical Queries
`
`Figure 1: Autonomous Scene Monitoring System.
`
`simply detect changes in· the incoming light 1 relying on operators to visually assess the
`situation.
`In principle, situation assessment can be performed a.utomatica.lly: by computers, to
`continuously exploit a.va.ilable CCTV surveillance data all of the time. Computer vision
`systems ca.n be used to extract information about the scene, such as how many people
`a.re in the field of view, and whether a particular kind of vehicle has entered the scene.
`In some cases this descriptive information may be all that is needed to facilitate efficient
`security monitoring and concise record keeping.
`
`2 Autonomous Scene Monitoring Architecture
`
`At Texas Instruments we have developed a. prototype scene monitoring system that au(cid:173)
`tomatically extracts complex information about scenes from CCTV camera. images and
`provides opera.tors with convenient access to the information.
`The overall scene monitoring system is illustrated in Figure 1. The Vision System
`detects people wa.lking in the CCTV camera field of view and continuously reports their
`3-D positions as they move. The Visual Memory a.t the c~nter of the scene monitoring
`system is an object-oriented data base that stores the information reported by the vision
`system. As people walk, their current 3·D positions are updated in visual memory, and a.
`history of their movement is maintained for future reference. The User Interface provides
`interactive graphical access to real-time and historical events stored in visual memory.
`The Vision System, Visual Memory, and User Interface software, are implemented
`on two computers. Th:e Vision System uses a Data.cube Ma.xVideo 20 real-time image
`processor and a. Sun SPARCstation 10, while the Visual Memory and User Interface run
`
`206
`
`

`

`on the Sun workstation. Users may also access Visual Memory over a network using an
`X Window System server or another workstation. The camera is a Texas Instruments
`monochrome MC-780PH CCD camera with a resolution of 755x484 8-bit pixels 1 of which
`492x460 pixels a.re used by the Vision Syste:i;n. The camera is mounted in a pan/tilt/zoom
`unit about 8 feet high on a hallway ce:iling in our laboratories. From this vantage point
`the camera can observe several hallways, including a section approximately 9 feet wide,
`117 feet long, and 12 feet high. The hallway is illuminated by overhead and wall-mounted
`fluorescent lighting, with significant variations caused by natural light from large windows
`at one end of the hall.
`
`3 Real-Time Scene Monitoring
`
`Algorithms of the Vision System report where people are walking in the field of view.
`The algorithms use basic image processing techniques to continuously detect scene mo(cid:173)
`tion. Regions of motion are analyzed for consistent interpretations as people standing or
`walking. By using knowledge of the scene geometry, the algorithms estimate positions
`and heights of people in the scene. The system detects and updates the positions of
`people at a rate of ten frames per second.
`
`4 Visual Memory
`
`Our visual memory prototype project developed an architecture [1] to interface vision
`systems with applications requiring access to informatio~ about 3-D objects, events and
`their environment. Visual Memory requirements include:
`
`• Storage of objects at high frame rates
`
`• Retrieval from multi-gigabyte data volumes
`
`• Support for diverse data structures
`
`We selected an object-oriented data base {OODB) (2 1 3) to use as the basis for the
`visual memory prototype. Relational data base technology is poorly suited for Visual
`Memory because it does not adequately support diverse data structures. The OODB is
`illustrated in Figure 2. For Visual Memory, the Indexing and Address Space modules of
`the OODB were extended to speed object storage and retrieval. In particular:, seve:ral
`spatio-temporal indexing mechanisms were introduced as described below·.
`
`4.1 Spatial Indexing
`
`Spatial indexing provides fast, efficient answers ~o questions such a.s, "Is anyone in area.
`X? 11 Spatial indices typically provide conservative approximate answers, allowing false
`positives but not false negatives, and rely on further :filtering for exact results.
`Several different spatial indices are available, catering to different types of questions.
`For example, grids and point quadtrees are good for determining objects near a given
`
`207
`
`

`

`Application
`
`Spatio-Temporal
`Extensions
`
`Meta Architecture Support
`
`Persistence Transaction Distribution
`
`Change
`
`Indexing
`
`Query
`
`Support Modules
`
`Address
`Space
`
`Communication
`
`Translation
`
`Data
`Dictionary
`
`Caching Extensions
`
`Figure 2: Object-Oriented Data Base.
`
`point~ while interval trees are better for finding object intersections [4]. The visual
`memory architecture supports multiple indices, and its query mechanism determines the
`appropriate index for a given question.
`
`4.2 Temporal Indexing
`
`Temporal indices provide means for accessing object histories by efficiently determining
`an object's state at a given time and how objects changed over a given interval of time.
`They help answer questions such as, "Was anyone in hall X during the night?,, and
`"Where was object" Y at 10:00am?"
`Mathematically, temporal indexing mechanisms may be regarded a.s special cases of
`spatial mdices for one dimension (time). Thus selection of a particular temporal index
`depends on the intended use. For example, image processing inputs to vis~al memory
`us~ only increasing time. Also, old informa.tion becomes increasingly unimportant and
`can be archived [5] for infrequent access. Tree indices support these requirements.
`
`5 User Interface
`
`The user interface provides interactive access to visuaj. memory data via historical and
`real-time queries. To make historical queries, the user specifies periods of time, regions
`of space, and object types. Then· the system retrieves the corresponding objects. To
`display real-time information, locations of people are indicated on a ·floor plan display
`and changed dynamically a.s the visual memory is updated.
`
`208
`
`

`

`Users may specify alarm regions on the map so that when someone enters that area
`a.n alarm is signaled. Digital snapshots of the scene may be kept and saved for future
`reference. Other user interface features provide interactive control of the camera pan,
`tilt, zoom, focus, and gain settings,
`
`6 Concluding Remarks
`
`The autonomous scene monitoring system currently operates in our laboratories with two
`remotely controlled pan/tilt/zoom cameras. We have also demonstrated operation of the
`system outdoors, using an infrared camera to map the position of a person walking along
`a sidewalk and across a street. The object-oriented data base architecture (3] underlying
`the Visual Memory facilitates expansion of the prototype scene monitoring system to
`handle hundr~ds of cameras, vision systems, and many users. As vision systems and
`visual memory evolve, increasingly sophisticated surveillance tasks will be automated to
`enhance security systems,
`
`Acknowledgments
`
`We thank Chris Kellogg, Steve Ford and Tom O'D«amnell for their contributions in the
`conception, design, and development of the autonomous scene monitoring system.
`
`References
`
`(l] Kellogg, C. 1 "Visual Memory", MIT Thesis, May 1993. Also available as: ~echnical
`Report CSL-93-05-20, Texas Instruments, Systems & Information Science Labora(cid:173)
`tory.
`
`[2] S. Ford, et a.I, '1ZEITGEIST: Database Support for Object-Oriented Programming,"
`Advances in Object-Oriented Database Systems, 2nd International, Workshop on
`Object-Oriented Database Systems, Springer-Verlag, Sep. 1988.
`
`[3] Wells, D.L., J.A. Blakeley, & C.W. Thompson, "Architecture of the Open Object(cid:173)
`Oriented Data.base Mana.ge~ent System," IEEE Computer, Vol. 25) No. iO, Oct.
`1992.
`
`[4] Samet, H., The Design and Analysis of Spatial. Data Structures. Addison-Wesley
`Publishing Company, Inc., 1989.
`
`[SJ Elmasri, R., M. Jaseemuddin, & V. Kouramajian, "Pa.rtioning of Time Index for
`Optical Disks," IEEE Data Engineering Conference, Feb. 1992.
`
`209
`
`