Descripción del proyecto
MOST OF TRAFFIC ACCIDENTS ARE CAUSED DUE TO DRIVER ERRORS OR LACK OF ATTENTION, BECAUSE OF THIS THEY ARE VERY DIFFICULT TO ERADICATE AS DRIVERS TEND TO NOT RECOGNIZE EASILY THEIR DEFICIENT DRIVING SKILLS, SOME PERSONAL LIMITATIONS LIKE DROWSINESS OR THEIR INSUFFICIENT ATTENTION, THE REDUCTION OF TRAFFIC FATALITIES IS A MAJOR CONCERN IN ADVANCED SOCIETIES AND ALTHOUGH MUCH HAS BEEN DONE WITH EDUCATIONAL CAMPAIGNS AND INFRASTRUCTURE IMPROVEMENTS, SOMETHING HAS TO BE DONE INSIDE THE VEHICLE THAT HAVE TO BE ABLE TO SENSE THEIR INTERNAL STATE AND THE ENVIRONMENT THAT SURROUNDS THEM AND MONITOR THE DRIVER,THIS LAST ASPECT IS THE GOAL OF THE VIDAS-DRIVER SUBPROJECT: TO MONITOR THE DRIVER STATE OF ATTENTION AND DROWSINESS USING COMPUTER VISION AND DEVELOP ARTIFICIAL CO-DRIVERS THAT WILL ASSIST HUMAN DRIVERS AS A WAY TO BUILD SAFER VEHICLES, IMAGE ANALYSIS OF THE DRIVER HAS THE MAIN ADVANTAGE OF MEASURING THE STATE OF THE DRIVER AND NOT THROUGH OTHER VEHICLE PARAMETERS LIKE THE PRESSURE ON THE STEERING-WHEEL OR THE DISTANCE TO THE ROAD LANE, THE MAIN CHALLENGES ARE: THE NEED OF A REAL-TIME SYSTEM, THE GREAT NUMBER OF DIFFERENT APPEARANCES, AND UNCONTROLLED ILLUMINATION, AN IMPORTANT REQUIREMENT IS THE HIGH ACCURACY, ALMOST 100%, AS DRIVER ARE ACCUSTOMED TO A NEARLY PERFECT MACHINE AND AN ELEMENT WITH DAILY FALSE POSITIVES WOULD BE A NUISANCE BUT ALSO A FALSE NEGATIVE COULD HAVE DEADLY CONSEQUENCES,IN THIS SUBPROJECT WE WILL DEVELOP TWO COMPUTER VISION SYSTEMS: ONE FOR THE FACE OF THE DRIVER AND THE SECOND FOR HIS/HER HANDS, WITH THE FIRST, THE 3D ORIENTATION OF THE HEAD OF THE DRIVER, THE GAZE ORIENTATION AND THE MOVEMENTS OF THE HEAD AND MOUTH WILL BE DETECTED, WITH THE SECOND, THE SYSTEM WILL KNOW WHETHER IF THE DRIVER HAS BOTH HANDS ON THE WHEEL, IS CHANGING GEARS OR PLAYING MUSIC, THE FUSION OF BOTH ANALYSES WILL ALLOW THE ATTENTION AND DROWSINESS OF THE DRIVER TO BE DETERMINED,A FREQUENT SHORTCOMING WITHIN THE RESEARCH IN THIS FIELD IS THE LACK OF A PUBLIC DRIVER IMAGE DATABASE FROM WHERE DIFFERENT APPROACHES CAN BE COMPARED, BECAUSE OF THIS, AN OBJECTIVE OF THE VIDAS-DRIVER SUBPROJECT IS THE CREATION OF THIS DATABASE, THIS IS BECAUSE SEVERAL APPROACHES NEED TO BE EXPLORED FROM MONOCULAR AND STEREO VISION TO 3D TIME-OF-FLIGHT CAMERAS, ALSO THE COMPARISON OF THE PROS AND CONS FOR EVERY CASE STUDIED, FOR ALL OF THEM, A 3D MODEL OF THE FACE WILL DETERMINE THE POSITION AND ORIENTATION OF THE FACE, THE NEXT STEP WILL BE TO FIND THE EYES AND OBTAINING THE GAZE DIRECTION, MOUTH MOVEMENT AND ANALYSIS WILL HELP TO KNOW IF THE DRIVER IS TALKING OR YAWNING, INSTEAD OF A SERIES OF INDEPENDENT STILL IMAGES, THE WHOLE SEQUENCES WILL BE ANALYZED AND THE TRACKING ALGORITHM WILL TAKE AN IMPORTANT ROLE IN THE RESEARCH, BESIDE THE DIFFICULTIES INHERENT TO THE TYPE OF OBJECTS THESE SYSTEMS HAVE TO RECOGNIZE, THE NATURAL ILLUMINATION WILL PLAY AN IMPORTANT CHALLENGE NOT ONLY BECAUSE THE SYSTEM WILL RUN UNDER REAL CONDITIONS AND THEREFORE THE RESULTS WILL BE TESTED UNDER SUNNY, CLOUDY, RAINY, DURING THE DAY AND DURING THE NIGHT, BUT ALSO BECAUSE HALF THE OBJECT CAN HAVE AN ILLUMINATION AND THE OTHER HALF CAN BE UNDER A SHADOW,THIS RESEARCH IS WITHIN THE PROJECT DENOMINATED CO-DRIVERS WHERE THE OTHER SUBPROJECTS ARE VIDAS-ROAD AND AIDA, THE FINAL GOAL IS TO MERGE BOTH THE INTERNAL AND EXTERNAL INFORMATION AND WARN THE DRIVER ONLY WHEN EITHER A DANGEROUS SITUATION IS GOING TO HAPPEN, HE IS NOT ANSWERING FAST ENOUGH OR IS NOT AWARE OF THE IMMINENT DANGER, DETECCION DE LA SOMNOLENCIA\MONITORIZACION DEL CONDUCTOR\SISTEMAS DE ASISTENCIA A LA CONDUCCION\VEHICULOS INTELIGENTES\VISION POR COMPUTADOR