UTOC

pancakesbootΤεχνίτη Νοημοσύνη και Ρομποτική

24 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

71 εμφανίσεις

© 2001 Mercury Computer Systems, Inc.

Underwater Tactical
Operations Center

(UTOC)


David A. Toms

Mercury Computer Systems

Dtoms@mc.com


© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

2

Outline


Trends in tactical C4ISR:


1991: TOCs can’t get enough intel support


2002: TOCs are inundated with intel products


Targeting timelines are increasing!


New tools are required to process the data


USAF/USA TOCs are migrating toward
mobile, lightweight, open architectures


Technology demonstrations at Mercury


Intelligent Bandwidth Data Compression


Aided Target Recognition


Multi
-
Hypothesis Target Tracking


Geo
-
registration


Open Wings


A new architecture for tactical op centers


© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

3



High
-
Value Targets

Military Convoys

Massed Forces


Function of Underwater Tactical Operations Center


Provide Surveillance Operator with Unified Tactical Picture via GIG


Tools to Review Sensor Information, Provide Contextual Information


Remove Sensor Clutter, Fuse Target Information

UTOC

Processing

Processing

Exploitation

(Sensor)

Tasking

Dissemination

Processing

Processing

Collection Platforms

‘Common Operational Picture’

UTOC ISR System

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

4

UTOC Processing


‘TPED Crisis’


Trained Image Analyst can
Process 10
-
30 Full
-
Scene
1Mega
-
pixel SAR Images/Hr
(Global Hawk Will Generate
~6,000 Images/Hour)


Trained Surveillance
Operator can Track 3
-
6
Targets (Joint STARS will
Generate 10,000 Target
Reports per second)


Trend


Semi and Fully Automated
Tracking, Registration, ATR,
Data Fusion


Semi
-
Automated Situation
Assessment, Sensor Tasking

TPED: Tasking, Processing, Exploitation, and Dissemination

PD: Probability of Detection, FAR: False Alarm Rate

Assisted Exploit.

ATR only

IA only

PD

FAR

How Well do IAs Classify Targets?

Dr. John M. Irvine, SAIC

Presentation at ATR Transition Conference, MIT Lincoln Lab,
June 7, 2000

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

5


Processing Algorithms

ATR

Fusion

Pattern

Analysis

Filtering

Detection

Situation

Assessment

Algorithm

Physics

Geometry

Statistical

Rule
-
Based

Model

Information

Processing

Knowledge

Processing

Intelligence

Type

Signal Processing

Data Processing

Processing

10
-

100

GFLOPS

100s of

GFLOPS

Throughput

Amount of Sensor Data (100s of MB/s)

Decreasing Amount of Data

Increasing Desired Information

Impact

Mercury

Domain

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

6

Signal and Data Processing
Architecture

MTI Signal

Processing

(e.g., STAP)

SAR Image

Formation

SAR

Registration

MTI Target

Detection

MTI

Registration

SAR Target

Detection

Video/IR

Registration

Target

Detection

SAR Feature

Extraction

Feature

Extraction

HRR Feature

Extraction

ATR

2D ATR

1D ATR

Tracking

Sensor

Tasking

Multi
-
Sensor/

Multi
-
Look

ATR Fusion

Chip
-
Level Processing

Image Processing

Report Processing

Signal Processing

Video/IR Signal

Radar

Signal

Sensor Mode/

Look Region

Image
-
Level Processing

‘Data’ Processing

Image Change

Detection

Fire Control

Systems

Signal

Analysis

Registration

ESM

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

7

UTOC Data Fusion

Types of Sensor Data to be fused:


Synthetic Aperture Radar imagery


Ground Moving Target Indicator reports


Electro
-
optic/Infrared imagery


Hyperspectral imagery


SIGINT/ELINT reports


COMINT data


BDA


Chem/Bio/WMD reports


Minefield delimitations


Unattended sensors

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

8

Stand
-
Alone Demonstrations


Tools are required to analyze all the
data


Completed demonstrations:


Intelligent Bandwidth Compression (Sandia
Labs/Black River Systems/Mercury)


Model
-
based ATR Algorithm (DARPA)


Geo
-
Registration (DARPA)


Under Development


Multiple Hypothesis Tracking (DARPA)

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

9

Intelligent Bandwidth Compression


Full
-
Scene Image
Compressed After Target
Chips Have Been Detected


High Ratio for Background


Low Ratio for Target Chips


Overall Ratio ~128:1

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

10

GMTI Tracking Function


Generate Target Tracks Based on
MTI Radar Reports


Tracker has to Account for


Non
-
constant Target Velocities


Measurement Errors


Missed Detections


False Reports

MTI REPORTS

TRACKS

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

11

GMTI Tracking Approach

PREDICT

TRACKS

PREPROCESS

REPORTS

GATE REPORTS

WITH TRACKS

UPDATE

TRACKS

TRACK

HYPOTHESIS

MANAGEMENT

MTI

Reports

DTED

Data

Road

Data

DTED

Data

Road

Data

Target

Tracks

Constrained

MTI Reports

Report/Track

Pairs

Updated Track

Hypotheses

Predicted Track

Hypotheses

Retained Track

Hypotheses

Time


Multitarget Tracking


Associate Reports for Tracks


Filter out Measurement Noise by Averaging


Multiple Hypothesis Approach (MHT)


Form Multiple Hypotheses for Report
Associations and Target Kinematics


Select Most Likely Hypothesis After Processing
Multiple Frames of MTI Reports

CREATE

NEW TRACKS

Search

Problem

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

12

Image Registration Function


Provides Spatial Correspondence
Between Two Images


Image Registration Prerequisite
for Performing Change Detection


Registration Results: Alan Chao, Alphatech

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

13

Image Geo
-
Registration Approach


Two
-
step Procedure


Image Geocoding (Image Ortho
-
rectification):
Image Projected to a Common Reference
Frame


Image Registration: Uses Image Data (Pixel
Intensities or Image Feature) to Find Spatial
Correspondence Between Images

Image

Geocoding

Image

Registration

Images

System

Error Statistics

Geocoded

Images

Geocoding

Error Statistics

Registered

Images

Registration

Error Statistics

Relative

Geometry

Search

Problem

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

14

Image Registration Requirements


Function of


Feature or Pixel Intensity Approaches


Features Used (e.g., Topological, Region, Object)


Matching Algorithm (e.g., Hausdorff Distance,
Bayesian Metric)


Desired Accuracy


Preliminary Estimates (Feature
-
Based
Algorithm for SAR Image Registration)


~450x350 Pixel SAR Image


10
-
100 Giga operations

Processing Requirements are Function of Input Rate, Data Characteristics, Desired Performance

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

15

ATR Function


Classify Target Based on Image Chip


2
-
D ATR: Uses High
-
Res SAR Image Data


1
-
D ATR: Uses High
-
Res Range Profile from
Radar MTI Data (Research Area)

T72

Confidence:0.95

2
-
D ATR

1
-
D ATR

SCUD TEL

Confidence:0.7

Range
-
Doppler

Target Chip

SAR

Target Chip

Range
-
Profile

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

16

2
-
D ATR Approach


Traditional Approach: Template
-
Based


Store Target SAR Templates for Various
Poses, Articulation


Find Best Match and Declare Target Type


Model
-
Based Approach (Research)


Store Wire
-
Frame Model for Various Target
Types


Predict, Evaluate, Match, Search for Best
Match of Pose and Articulation

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

17

2
-
D ATR Requirements


Preliminary Estimates


From Demo System


Chip
-
Level Parallel Processing








Approx. 40 Mega operations per Chip

Ground Truth

System call &
score, if
correct

System call &
score, if
incorrect

T72

T72

BTR70

SA8

ZSU

T72

BTR70

ZSU

SCUD

T72

T72

BTR70

ZSU

T72

SCUD

SA8

ZSU

BTR70

ZSU

BTR70

SA8

ZSU

T72

T72

BTR70

SA8

ZSU

T72

BTR70

ZSU

T72

T72

T72

BTR70

SA8

ZSU

T72

BTR70

SCUD

.95

.90

.85

.80

.95

.95

.65

.80

.90

Click on image
chip to inspect
ATR details

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

18

Model
-
Based ATR Algorithm

Type (x,y)

†††
Sc潲e

M2

35,87


10

0.91

M548

38,88


14

0.83

BMP2

32,89

192

0.05

...

Focus of

Attention

Index

Search

Predict

Match

Extract

SAR Image

Detect

Scene Hypothesis

ROIs

ROI

Predicted

Features

Extracted

Features

Evaluations

Scene Model

Coarse

Hypotheses

Cue

Explain

Verify

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

19

Mercury Role in UTOC


Established Leader in Signal Processing


Expertise in Sensor Algorithm Technology


Middleware to Support Application Development


Low space, weight and power requirements


Mercury HPC Architecture Well
-
Suited
for Data Processing


Can be Scaled to Support Improvements in Sensor
Resolution


Supports Algorithms Requiring Tight Coupling
Between Signal & Data Processing Functions


Testing underway at WPAFB Sensor laboratory

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

20

Participation in Ground Station
Activities: Openwings

Openwings

Architecture

HPC

UTOC

Processing

Requirements


Openwings:
Architecture for Plug
-
and
-
Play, Network
Centric, Service
Oriented System


Mobile Ground Stations
is a Domain Example


Mercury is on Expert
Team


Analyzing Ground Station
Processing Requirements


Developing HPC Container

Specifications


Life Cycle Support for an
Application


Clustering of Processors


Process Load Balancing

www.openwings.org

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

21

Program Overview


Openwings initiative established June
‘99 as a Joint IR&D effort between
Motorola & Sun


Motorola and Sun Microsystems to lead
a community in the development of a
distributed, self
-
forming architecture


Mercury will provide High Performance
Computing engines


Architecture development will be done
using an open development approach


Initial framework is available to the
Openwings community

© 2001 Mercury Computer Systems, Inc.

Ground Stations 1.5

22

Summary


If submarines are to become full players
in network centric warfare, then
accessing and exploiting all available
data sources will become essential


We are performing groundwork to show
Mercury’s computers can meet these
requirements


Conducting data processing requirements analysis
and preparing demonstrations


These tools could be used for sonar data
exploitation as well.