Download Los Alamos National Laboratory

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Transcript
Adaptive Multi-sensor
Integrated Security System
(AMISS)
August 1998
AMISS Agenda
•
•
•
•
•
•
•
•
AMISS Overview
Source Detection and Location
Path Tracking Anomaly Detection
Multi-sensor Data Fusion
Shape Recognition and Identification
Reasoning/Data Mining
Remote Operation
AMISS Integrated Exercise
Los Alamos National Laboratory
2
AMISS Vision
Develop and demonstrate state-of-science
adaptive technology to determine potential
threats and anomalous situations and ascertain
appropriate action to increase facility security,
safeguards, and safety.
Los Alamos National Laboratory
3
AMISS Strategic Objectives
Los Alamos National Laboratory
4
AMISS Strategic Objectives
•
•
•
•
•
Provide continuous, adaptive real-time
detection and categorization of all activity
Assist security personnel
Enhance material movement monitoring
Reconstruct threatening events
Ensure compliance
Los Alamos National Laboratory
5
AMISS Tactics
•
•
•
•
•
Develop test bed
Learn correct/acceptable facility operating
conditions and detect unusual behavior
Build techniques for active facility
Build portable capabilities
Provide feedback to security personnel
Los Alamos National Laboratory
6
AMISS Challenge
Los Alamos National Laboratory
7
Los Alamos National Laboratory
8
AMISS Architecture
Whole Is Greater Than The Sum Of The Parts
Los Alamos National Laboratory
9
AMISS Schematic
Los Alamos National Laboratory
1
0
AMISS Architecture
Los Alamos National Laboratory
11
Source Detection And Location
•
Long-term vision
– Detect, identify, and locate multiple sources
and track movement and activity
•
Initial goal
– Detect and locate single moving source in a
room
Los Alamos National Laboratory
1
2
Source Location Problem
•
•
•
•
•
•
Detector gives imperfect information
Source moves around room
Inconsistent signals (candle flickers)
Background changes
Detectors are non-directional
Shielding or occlusions
Los Alamos National Laboratory
1
3
Source Location At A Glance
•
Isocounts as single bands with known source
strength
Los Alamos National Laboratory
1
4
Simplistic Source Location
•
Overlapping isocount Bands gives X,Y,Z and
strength
Los Alamos National Laboratory
1
5
Source Location Solution
•
•
•
•
Estimate X, Y, Z background levels
Radiation detection model for rectangular
detector and spherical source
Optimization algorithm
– Kalman Filter
– Constrained optimization
Elements of uncertainty
–
–
–
–
Counting statistics
Occlusion, shielding
Varying background
Directional detectors
Los Alamos National Laboratory
1
6
Source Location Future
•
•
•
•
•
Improve Location Accuracy with known
source strength
Locate Multiple Sources
Deduce multiple source strength
Deduce Nuclear Signatures
Decrease uncertainties
– Location
– Source Strength
– Source Material
Los Alamos National Laboratory
1
7
AMISS Architecture
Los Alamos National Laboratory
1
8
Path Tracking Anomaly Detection
•
Automated process using data collection to
locate individual movement to detect and
determine anomalous behavior patterns
Los Alamos National Laboratory
1
9
Path Tracking Anomaly Detection
Current Technology
•
•
•
•
Unsupervised neural network
Clusters spatial-temporal patterns
Learns individual behavior
Defines normal behavior
– X, Y locations in sequence
– Broader patterns
Los Alamos National Laboratory
2
0
Path Tracking Anomaly Detection
Los Alamos National Laboratory
2
1
Path Tracking Anomaly Detection
Los Alamos National Laboratory
2
2
Path Tracking Anomaly Detection
AMISS Innovative Technology
•
•
•
•
•
Determine anomalies from examples of
normal behavior - more secure
Real-time
Provide explanation facility
Learns quickly
Customized by changing sensitivity
parameters
Los Alamos National Laboratory
2
3
Path Tracking Anomaly Detection
Future
•
Facility Security
–
–
–
–
•
Identify number of people in a room
Identify open doors
Examine time
Determine expected behavior
– Expand to larger areas
Other domains
– Verification of dismantlement activities
– Assist IAEA inspections
– Ensure unattended monitoring
Los Alamos National Laboratory
2
4
AMISS Architecture
Los Alamos National Laboratory
2
5
Multi-sensor Data Fusion
•
Combining multi-sensor data to provide more
accurate, descriptive, and useful information
for higher-level reasoning and display
Los Alamos National Laboratory
2
6
Multi-sensor Data Fusion
Challenges
•
Different sensor characteristics
–
–
–
–
–
Different data collected
Various accuracy
Various reliability
Different resolutions
Varying speed
Los Alamos National Laboratory
2
7
Multi-sensor Data Fusion
Current Research
•
•
•
•
•
Uses proven graph theory technique,
including edge trimming
Robust sensor suite
– Handle sensor failure
– Redundancy for level of assurance
Fuse active and passive infrared and video to
identify and locate personnel
Experimental code complete
Evaluation phase starting
Los Alamos National Laboratory
2
8
Multi-sensor Data Fusion Future
•
•
•
•
Integrate into AMISS
Research other methodologies
Add new sensor types
Framework for future development
Los Alamos National Laboratory
2
9
AMISS Architecture
Los Alamos National Laboratory
3
0
Shape Recognition and
Identification - Initial Concept
•
Profile Evaluator for image identification
within controlled environment for entry
control
Los Alamos National Laboratory
3
1
Shape Recognition and
Identification - Current Technology
moment p,q = 
x,y
x p y q f(x,y)
Los Alamos National Laboratory
3
2
Shape Recognition and
Identification - Current Technology
•
Neural Net
y= Wi Xi
z = tanh (y)
Los Alamos National Laboratory
3
3
Shape Recognition and
Identification - Current Results
ERROR RATES
Actual
Estimated
1 Camera
3 Cameras
Tailgating
.1700
.0049
Crouching
.0000
.0000
Person +
Object
.0625
.0002
Normal
.0870
.0007
Los Alamos National Laboratory
3
4
Shape Recognition and
Identification
Los Alamos National Laboratory
3
5
Shape Recognition and
Identification Future
Los Alamos National Laboratory
3
6
Shape Recognition and
Identification - Future
•
Detect and identify a complete inventory of
objects in a scene
– New Feature Extraction - Image Understanding
– Better hardware
– Post processing classification
Los Alamos National Laboratory
3
7
AMISS Architecture
Los Alamos National Laboratory
3
8
Reasoning/Data Mining
•
•
ADaptive Virtual Integrating senSOR
(ADVISOR)
Provide continuous, integrated facility status
reasoned from real-time and historical data
and human input.
Los Alamos National Laboratory
3
9
ADVISOR Expertise
Los Alamos National Laboratory
4
0
ADVISOR Benefits
•
•
•
•
•
•
•
Provides continuous real-time detection of
procedure violations and anomalies
Reduce information overload and tedious
data analysis
Consistent rule interpretation and application
Continuous information integration
Continuity of knowledge
Provides detailed explanation capability
Active role, advisory role or combination
Los Alamos National Laboratory
4
1
ADVISOR Objectives
•
•
•
•
Knowledge engineering to obtain human
expert experience
Integrate policies and procedures
Dynamic adaptation
Learn normal facility status
Los Alamos National Laboratory
4
2
ADVISOR Components
•
•
•
Real-time reasoning and control (current)
Data mining
Real-time integrated with data mining
Los Alamos National Laboratory
4
3
ADVISOR Decisions
Los Alamos National Laboratory
4
4
ADVISOR Decision
YES
Is Joe supposed to
move that material
in that location?
Los Alamos National Laboratory
4
5
ADVISOR Future
•
•
•
•
•
Develop System Health reasoning diagnostics
Expand to multiple buildings and facilities
Countermeasures (e.g. know when being
fooled)
New domain applications
Develop and integrate data mining
Los Alamos National Laboratory
4
6
AMISS Architecture
Los Alamos National Laboratory
4
7
Remote Operation
•
•
•
•
World Wide Web (www)
– Remote (e.g. global)
– Secure (SSL)
Remote alarm (e.g. notify guard through
pager)
Potential to take action immediately
Operate several facilities
Los Alamos National Laboratory
4
8
Remote Operation Capabilities
•
Bringing it all together - remotely
ACTION
Shut Door
ADVICE/STATUS
All OK
CONFIRM ACTION
Dispatch Guard
QUERY HUMAN
EXPERT
Human hint to confirm
ADVISOR Thought
Los Alamos National Laboratory
4
9
AMISS
Los Alamos National Laboratory
5
0
Potential Application Areas
•MC&A at DOE Facilities
•Treaty Verification
•IAEA
– Remote (unattended) monitoring
– Environmental monitoring
– Covert and/or underground facilities
•Other
– Critical infrastructure
– Recent national and international incidents
Los Alamos National Laboratory
5
1
AMISS Future
•
•
•
•
•
•
•
•
Confidence versus redundancy
Aging facilities
Robust and hardened
Adaptable, portable, scalable
Address emerging requirements
Define measurable performance measures
Further reasoning development
Integrate to entire facility
Los Alamos National Laboratory
5
2
AMISS Experimental Site — TA-18
Los Alamos National Laboratory
5
3
AMISS Exercise - Conceptual Demo
Los Alamos National Laboratory
5
4
AMISS Exercise - A Day In The Life
•
•
•
•
•
Identify and locate personnel and material
Alert unauthorized activities, material
shielding, and unauthorized movements
Track paths of interest and detect anomalous
behaviors
Provide real-time facility status decisions
based upon all available data sources
Provide security personnel with effortless
facility status view
Los Alamos National Laboratory
5
5
AMISS Exercise - A Day In The Life
1. Unlock High Bay
– Alarm unauthorized entry
2. Security Sweep High Bay
– Alert attempted sweep against protocol pattern
3. Experiment Entry Procedures
– Alert improper approvals, personnel, materials,
or schedule
Los Alamos National Laboratory
5
6
AMISS Exercise - A Day In The Life
4. Perform Experiment
– Alert improper procedures with material
– Alarm invalid exit while material shielded
5. Experiment Exit Procedures
– Alarm unauthorized material removal
6. Secure High Bay
Los Alamos National Laboratory
5
7
AMISS Integrated Exercise
Los Alamos National Laboratory
5
8
Partnering To Solve Problems
NN-20
Advanced Technology
Push State-Of-The-Science
Identify Gaps & Fill Them
Provide Vision/Path Forward
Leverage Opportunities
Anticipate Customer Needs
Provide Alternate Methods To Address
Current & Emerging Challenges
NN-40
Define Needs
Identify Challenges
NN-50
Define Needs
Identify Challenges
Los Alamos National Laboratory
5
9
Differences from Radiation
Instrumentation Problems
•
Instrumentation imperfections
– Moving source within integration interval
– Sensor sensitivity decreases with angle and
distance
– Minimal detector technology
Los Alamos National Laboratory
6
0
Radiation Detection
•
•
Identify “Unusual” background changes
Factors in developing “Decision Rules”
–
–
–
–
Recognizable source strength
Recognition time
Time between events
Acceptable error rates
•
•
False positive
False negative
Los Alamos National Laboratory
6
1
Radiation Detectors
•
Gamma Detector
– Gamma hits cause Fluorescence
– Light flashes counted
Los Alamos National Laboratory
6
2
THE AMISS TOOLBOX
each component has equal status and every component can communicate with
every other component
Sensors
video
camera
Virtual
Sensors
Expert
Systems
Anomaly
Detection
Data
Mining
Brain
Interface
expert
system
demo
site
specific
path
tracking
DOE
rules
PTAD
PIRs
object
recognition
LANL
rules
multiple
people
personel
profiles
neural
network
palm
reader
source
location
site specific
rules
source
SNM
profiles
mixture of
experts
AIRs
multiple
sensor
fusion
safety
rules
event
driven
sensor
profiles
case-based
reasoning
database
Los Alamos National Laboratory
Component Communication
dialogue
security
6
3
AMISS COMPONENT COMMUNICATION
each component contains the same communication software
PTAD Specific Code
Anomaly Detection
tracker
hot_spot
PTAD
Communication Code
multiple
people
C_API
dataformat
source
source Specific Code
source
event
driven
variance
Communication Code
C_API
dataformat
Los Alamos National Laboratory
6
4
DEMO FLOW DIAGRAM
Web Interface
Expert System
PTAD
Blob Tracking
Rad Tracker
Sentry
Face It
PIR/AIR
Portal Monitor
Video
Bar Code
Los Alamos National Laboratory
Rad Detectors
Palm Reader
Database
6
5
AIR
INTERFACE
PIR
Web Browser
ROOM STATE
Expert System
ACCESS CONTROL
PEOPLE STATE
SOURCE STATE
Sentry
PTAD
Face It
Rad Tracker
Data Base
MOTION DETECTION
Palm Reader
Video Tracking
Rad
Detectors
Bar Code
Reader
Los Alamos National Laboratory
Video
6
6
Document related concepts
no text concepts found
Similar
Ellen Cerreta Trustee-Elect (2015-2018)
Ellen Cerreta Trustee-Elect (2015-2018)