smc platform and adaboost face tracking

29
SMC Platform and Adaboost Face Tracking • Introduction to SMC system arch itecture • Introduction to Ethernet protoc ol • Implementation on SMC board • Adaboost Face Tracking Algorith m and application

Upload: watson

Post on 11-Jan-2016

54 views

Category:

Documents


4 download

DESCRIPTION

SMC Platform and Adaboost Face Tracking. Introduction to SMC system architecture Introduction to Ethernet protocol Implementation on SMC board Adaboost Face Tracking Algorithm and application. SMC Platform. Target Board (SMC) 傳輸影像到 PC by Ethernet. (SMC). 同軸電纜線. camera. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: SMC Platform and Adaboost Face Tracking

SMC Platform and Adaboost Face Tracking

• Introduction to SMC system architecture

• Introduction to Ethernet protocol

• Implementation on SMC board

• Adaboost Face Tracking Algorithm and application

Page 2: SMC Platform and Adaboost Face Tracking

SMC Platform

同軸電纜線

Target Board (SMC)傳輸影像到 PC by Ethernet

(SMC)

camera

Page 3: SMC Platform and Adaboost Face Tracking

SMCDigital Camera SEEDJTAG

PC(CCS)

USB/PCI

PC(Client)

Ethernet

BNC

RS485&RS232

Page 4: SMC Platform and Adaboost Face Tracking
Page 5: SMC Platform and Adaboost Face Tracking

Composite video(CVBS)

• Format of analog television (picture only)

• Composite of Y U V.

• Standard format– NTSC (525-line/60Hz)– PAL (625-line/50Hz ) ♪

Page 6: SMC Platform and Adaboost Face Tracking

RAC

Page 7: SMC Platform and Adaboost Face Tracking

BNC

Page 8: SMC Platform and Adaboost Face Tracking

F Connector

Page 9: SMC Platform and Adaboost Face Tracking

S-Video (S-VHS)

• Separate Y , C signal.

• Simple to convert S-Video to composite signal.                                                        

Pin 1 GND Ground (Y)

Pin 2 GND Ground (C)

Pin 3 Y Intensity (Luminance)

Pin 4 C Color (Chrominance)

Page 10: SMC Platform and Adaboost Face Tracking

VGA(Video Graphics Array)

• Maximum of 720 horizontal pixels

• Maximum of 480 lines

• Refresh rates at up to 70 Hz

Page 11: SMC Platform and Adaboost Face Tracking

RS232 vs RS485

SPECIFICATIONS RS232 RS485

Mode of Operation

SINGLE-ENDED

DIFFERENTIAL

Maximum Cable Length

50 FT. 4000 FT.

Page 12: SMC Platform and Adaboost Face Tracking

JTAG( Joint Test Action Group )

• Formed in 1985 to develop a method to test populated circuit boards after manufacture.

• Primarily used for accessing sub-blocks of integrated circuits, and is also useful as a mechanism for debugging embedded systems.

Page 13: SMC Platform and Adaboost Face Tracking

SMC Platform

Page 14: SMC Platform and Adaboost Face Tracking

Ethernet protocol

• OSI 網路

Page 15: SMC Platform and Adaboost Face Tracking

Ethernet protocol

• IEEE802.3/Ethernet MAC frame format

Page 16: SMC Platform and Adaboost Face Tracking

Ethernet connection

Texas Instrument

TMS320DM642 DSP

MII

Intel LXT972A

Page 17: SMC Platform and Adaboost Face Tracking

IEEE 802.3 CSMA/CD

• Carrier Sense Multiple Access with Collision DetectionFrame

TransmitterFrameReceiver

TransmitFrame

TransmitDataEncap

ComputePad CRC32

TransmitLinkMgmt

WatchForCollision BackOff

StartTransmit Random

BitTransmitter

PhysicalEncap

StartJam NextBit

TransmitBit Wait

ReceiveFame

ReceiveDataDecap

RecognizeAddress RemovePad

ReceiveLinkMgmt

StartReceive

Deference

RealTimeDelay

BitReceiver

PhysicalDecap

ReceiveBit

MEDIA ACCESS SUBLAYER

LLC SUBLAYER

PHYSICAL LATER

TRANSMITRECEIV

E

MEDIUMMANAGEMENT

FRAMING

Page 18: SMC Platform and Adaboost Face Tracking

IEEE 802.3 CSMA/CD

MAC Sublayer 5 processes• Frame transmit process (Data encapsulation & Transmit media

access management )• Bit transmit process• Bit receive process• Frame receive process (Data decapsulation & Receive media a

ccess management )• Deference process (Half-duplex mode & Full-duplex mode )

Flow Control -> Pause Frame

Page 19: SMC Platform and Adaboost Face Tracking

• Transmit frame process TransmitFrame

TransmitENABLE?

assemble frame

defferringon?

start transmission

collisionDetect?

transmissionDone?

too many attempts?

incrementattempts

send jam

computbackoff

wait backofftime

Done:excessiveCollisionError

Done:transmitOK

Done:transmitDisabled

no

yes

no

yes

no

yes

no

yes

yes

no

Page 20: SMC Platform and Adaboost Face Tracking

• Receive frame process

ReceiveFrame

ReceiveEnable

StartReceiving

DoneReceiving

RecognizeAddress

Frame tooLong

ExtraBits

Disassemble Frame

Done:ReceiveOK

Done:Length Error

Done:Frame too Long

no

no

no

yes

yes

yes

Frame too Small

Valid FrameCheck

Sequence

Valid Length

Done:Alignment Error

Done:Frame Check Error

Done:ReceiveDisabled

no

no

no

non

o

yes

yes

yes

yes

yes

Page 21: SMC Platform and Adaboost Face Tracking

• Deference process

ChannelBusy

channel free

frame wainting

no

no

no

yes

yes

yes

defferring on

deferring off

Page 22: SMC Platform and Adaboost Face Tracking

Ethernet protocol

• Bit transmit process and bit receive process

transmission started

end of frame

no

no

yes

yes

transmit a bit

transmission done

receiving started

carrier sense on

no

no

yes

yes

receive a bit

receiving done

Page 23: SMC Platform and Adaboost Face Tracking

Ethernet protocol

• Ethernet MAC Module Block Diagram

Page 24: SMC Platform and Adaboost Face Tracking

Ethernet protocol

MII (Media Independent Interface)• Independent Interface Enable to use different mediu

m of Physical layer • The linkage between MAC layer and Physical layer • The linkage between Physical layer and Station managem

ent (MDIO)• The synchronization of data transmit and the clock period

of transmit and receiveMDIO (Management Data Input /Output)• The communication between Physical layer and Station m

anagement

Page 25: SMC Platform and Adaboost Face Tracking

Find Adaboost Face Tracking code

Test performance using picture or

Web Cam

Analyze Source Code

Test performance using Camera

Edit Code on VCPort to CCS (Data from computer)

Port to CCS (DATA from

Camera)Speed Up

Combine with Ethernet

&RS232

Implementation on SMC board

Page 26: SMC Platform and Adaboost Face Tracking

Program Source

• Georgia Institute of Technology

• Code released at 2005

• Contain multiple training method

• Contain trained data based• Fast Asymmetric Learning for Cascade Face Detection-J Wu, SC Br

ubaker, MD Mullin, JM Rehg - 2005

Page 27: SMC Platform and Adaboost Face Tracking

Progress

• Remove MFC (Window GUI )

• Remove learning part

• Remove OpenCV

• Remove iostream (132KB)

• Remove c++ vector

Page 28: SMC Platform and Adaboost Face Tracking
Page 29: SMC Platform and Adaboost Face Tracking