8/13/2019 An EPFL Student Has Developed a Video Analysis Algorithm Able to Estimate the Level of a Driver

1/3

An EPFL student has developed a video analysis algorithm able to estimate the level of a driver's

fatigue based on the degree of eyelid closure. PSA Peugeot Citron a pro!ect partner has built a

prototype in order to test it in real driving conditions.

"early a third of high#ay accidents are caused by fatigue. "o#adays there e$ist several

attention detection systems for drivers e.g.% based on the loss of vehicle control. &arinaimmermann as part of her master thesis in electrical and electronic engineeringand in

partnership #ith PSA Peugeot Citron has focused directly on drivers and their state of#a(efulness.She developed an algorithm able to measure the degree of eyelid closure by using a

single camera and thus avoidinginvasive methods.

)ne of the indicators most fre*uently used to determine the state of a driver's fatigue is+PE,CL)S+ -percentage of eye closure. /t measures the percentage of time that the pupil is at

least 012 covered by the eyelid during a predetermined timespan. 3he challenge #as to measure

this indicator in real time (eeping the drivers' eyes on the vie#finderand bearing in mind that

they can turn their head #ear glasses drive during the night cross through tunnels etc.

3he Signal Processing Laboratory 4 -L3S4 at EPFL is speciali5ed in face recognition

monitoring and analysis techni*ues and has man6machine interfaces capable of (eeping trac( of

a driver's face in real time. 3his is achieved by placing a small infrared camerabehind the #heel.

/n order to carry out the pro!ect 6initiated by the 3ransportation Center and carried out at L3S46&arina immermann developed an eye6analy5ing module by creating an algorithm able to

disregard possible light effects as #ell as the drivers' different eye morphologies. She then

established a 78 profiling of the eye and eyelids so as to distinguish an open one eye from aclosed one. Lastly she had to optimi5e her methodology to ma(e it possible for it to run in on6

board vehicle computers #ith limited computing po#er in real6time conditions.

78 representation of an opened and closed eye 6 the detector is able to interprete them

Industrialization potential

3he first tests carried out under real conditions sho#ed good reliability. 9o#ever the system

remains highly dependent on the frame rate as unconscious blin(s of the eye occur #ithin :11 to

http://phys.org/tags/electronic+engineering/http://phys.org/tags/wakefulness/http://phys.org/tags/wakefulness/http://phys.org/tags/invasive+methods/http://phys.org/tags/invasive+methods/http://phys.org/tags/invasive+methods/http://phys.org/tags/viewfinder/http://phys.org/tags/face+recognition/http://phys.org/tags/infrared+camera/http://phys.org/tags/wakefulness/http://phys.org/tags/invasive+methods/http://phys.org/tags/viewfinder/http://phys.org/tags/face+recognition/http://phys.org/tags/infrared+camera/http://phys.org/tags/electronic+engineering/

8/13/2019 An EPFL Student Has Developed a Video Analysis Algorithm Able to Estimate the Level of a Driver

2/3

:41 milliseconds. +3he proposed algorithm is sufficiently robust and simple to run on a standard

camera. /t #ill be able to combine the degree of eyelid6opening information together #ith other

data 6 li(e ya#ning or head tilt6 provided by the already e$isting face trac(ing system +said;ean6Philippe 3hiran head of L3S4 and supervisor of the student's master pro!ect.

As it stands the pro!ect's industriali5ation potential has already caught the eye of PSA PeugeotCitron pro!ect partner #hich #ill integrate the device into a first stage test vehicle for

evaluating its performance in real conditions.

-Phys.org

8/13/2019 An EPFL Student Has Developed a Video Analysis Algorithm Able to Estimate the Level of a Driver

3/3

camera. /t cannot obstruct the driver's field of vision. +)ne of the possible options is to place it

behind the steering #heel #hich #ould re*uire a system that's *uic( enough to recapture the

face detection after the arms of the steering #heel interrupt it+ e$plains 3hiran. Another issue isadapting to changing lighting conditions in order to trac( the driver's face as the car goes into a

tunnel for e$ample or #hen driving into a setting sun. And finally the tool must perform as

#ell #hen the driver is facing it as #hen she turns her head to the side.

Improving performance

+@e are currently #or(ing on improving the system's performance particularly by increasing the

number of images processed+ continues 3hiran. +3he ne$t step is to test it in realistic conditions.

PSA Peugeot Citron #ill install the platform on a car and our lab #ill test the system undervarious application scenarios.+

PSA Peugeot Citron opened an office in EPFL's /nnovation S*uare in summer B1:: #ith the

goal of collaborating on research and innovation pro!ects. Spearheaded by EPFL's 3ransportation

Center other pro!ects #ith EPFL laboratories have been launched including energy systemsdriver assistance systems human6machine interfaces and predictions for the structure of the

automotive mar(et.

The propose system based on eyes closer count & yawning count of the driver. Bymonitoring the eyes and mouth, it is believed that the symptoms of driver fatigue can bedetected early enough to avoid a car accident. The eye blink frequency increases beyondthe normal rate in the fatigued state.