You are on page 1of 16

OUTLINES

Requirement of driver vigilance system


Review of previous study
General system architecture
Experimental result Experimental result
Conclusion
REQUIREMENT
In europe,statistics show that 10% to 20% accidents are
caused due to driver with less vigilence level.
It is caused by fatigue,distraction.
It involves single vehicle and sober driver travelling alone.
As assistant system for comforts,navigation increase inside
car,distraction from more basic task-driving is observed.
PREVIOUS STUDY
MIT SMART CAR
ASV(ADVANCE SAFETY VEHICLE)BY TOYOTA
TCD(TECH CO DRIVER)SPANISH SYSTEM
ALL ABOVE MENTION SYSTEMS ARE INTRUTIVE-AS
THEY REQUIRE ELECTRODES TO BE ATTACHED TO THEY REQUIRE ELECTRODES TO BE ATTACHED TO
THE DRIVER CAUSING ANNOYANCE.
DAIMLER CHARYSLER SYSTEM BASED ON VEHICLE
SPEED,STEARING ANGLE,VEHICLE POSITION
RELATIVE TO ROAD DIMENSION.
MEASURE OF REDUCED ALERTNESS
Longer blink duration
Slow eyelid movement
Smaller degree of eye opening
Frequent nodding
Yawning Yawning
Gaze(narrowness in line of signal)
Sluggish facial expression
Computer vision can be natural and non intrutive for
extracting visual characteristics.
WHAT SHOULD WE LEARN FROM
PRIVIOUS STUDY?
There must be multiple clue to indicate drivers alertness.
System should be non intruitive
System should work satisfactorily at night also.(at low light
condition)
Use infrared led ring Use infrared led ring
Focus more on pupil detection and tracking.
SYSTEM ARCHITECTURE
MAJOR MODULES:
Image acqusition
Pupil detection
Visual behaviour Visual behaviour
Driver inattentiveness indicator software
IMAGE ACQUISITION
IR leds are used to brighten the users face
IT produce bright pupil effect
LEDS are mounted around camera axis.
Ring size is choosen such that inner ring produce bright pupil
effect while outer ring produce dark pupil effect. effect while outer ring produce dark pupil effect.
Get difference bet. two image and get pupil.
PUPIL DETECTION AND TRACKING
PUPIL DETECTION AND TRACKING
Difference dark pupil and bright pupil image
So bright blob in image can be detected by an adaptive
threshold.
Tracking of pupil is done by two kalman filters.
Kalman filter fail when sudden head movement. Kalman filter fail when sudden head movement.
It can be overcame by adaptive window.
PUPIL DETECTION AND TRACKING
VISUAL BEHAVIOUR
EYELID MOVEMENT
EYE CLOSURE DURATION
BLINK FREQUENCY
FIXED GAZE
EYE CLOSER/OPENING SPEED EYE CLOSER/OPENING SPEED
PERCLOS(PERCENTAGE CLOSERE)
FACE POSE
DEGREE OF EYE OPENING IS CHARACTERIZED BY PUPIL
SHAPE
AS EYES CLOSE,THE PUPILS START GETTING OCCLUDED
BY EYELID THIRE SHAPE GET MORE ELLIPTICAL.
DRIVER VEGILENCE COMPUTATION
WE CAN USE SOFTWARE WE DECIDE DIL(DRIVER
INATTENTIVENESS LEVEL) BY CONSIDERING ALL ABOVE
MENTION POINTS
SUCH THET
IF PERCLOS IS SMALL AND EYE CLOSURE DURATION IS IF PERCLOS IS SMALL AND EYE CLOSURE DURATION IS
SMALL AND FACE POSITION IS MEDIUM AND NODDING
FREQUENCY IS SMALL AND BLINK FREQUENCY IS MEDIUM
AND FIXED GAZE IS SMALL, THEN DIL IS SMALL.
IF PERCLOS IS MEDIUM LARGE AND EYE CLOSURE
DURATION IS MEDIUM AND BLINK FREQUENCY IS
MEDIUM AND FIXED GAZE IS LARGE, THEN DIL IS LARGE.
FINAL SYSTEM BLOCK DIAGRAM
CONCLUSION
WE DEVELOPED NONINTRUITIVE SYSTEM FOR
DRIVER VIGILENCE.
WE FOCUS ON SIX PARAMETERS PERCLOS,EYE
CLOSURE DURATION, BLINK FREQUENCY, NODDING
FREQUENCY, FACE POSITION, AND FIXED GAZE. FREQUENCY, FACE POSITION, AND FIXED GAZE.
SYSTEM WORKS SATISFACTORILY AT MIDNIGHT AND
DURING LESS LIGHT.
IT PERFORMANCE IS DECREASE DURING DAY TIME AND
PERSON WITH SUNGLASSES.
Assignment Assignment Assignment Assignment [Weightage 10%] [Weightage 10%] [Weightage 10%] [Weightage 10%]
Task 1: last date for Task-1 submission: on or before 05-10-2013 [batch wise]
Design a driver vigilance system.
Write the specification of the system.
Develop a system level model in C language as it is a universal language.
Task 2: [will be announced later] Task 2: [will be announced later]
Estimate the performance of the code on PowerPC.(Do profiling)
Estimate which part of the design should be migrated to HW on the
FPGA.
Interface the migrated design to bus.
Write the driver/software.
Document the result.
Input : set of images of the driver

You might also like