The FRAV research group organized its annual research meeting at the San Clemente Cultural Centre in Toledo. It is a perfect occasion to improve the internal organization of the group, as well as to plan the research work for the new academical year 2011/2012.
Currently FRAV is composed by four doctors, three PhD students and three researchers contracted for specific projects. Our research lines are: face recognition, computer vision for traffic and road safety, and computer vision for airport security.
Since today, the Face Recognition & Artificial Vision (FRAV) research group from the Universidad Rey Juan Carlos is also on the Facebook social network. Those people interested in the activities performed by FRAV will be able to become fans if they have their own Facebook account. They will receive easily news updates, information about our popularization activities or the scientific work done by our group.
To access our Facebook page, you just have to click the Facebook icon on the top right corner of this website, and then, just click the "I like" button. The subscription to FRAV's Facebook page is totally free and all the contents is open. The articles will be written in Spanish.
PhD Enrique Cabello, principal investigator of FRAV group (Face Recognition and Artificial Vision), and M.Sc. Óscar S. Siordia, FRAV researcher, have appeared today on Antena 3 TV news at 9 pm. Both have shown the work done by FRAV for CABINTEC project, in particular our automatic hand detector for drivers, which is specialized in truck drivers. This system tracks the position of the hands and allows detecting an inadequate driving behaviour, such as using the radio or a mobile phone or driving without holding the steering-wheel.
From FRAV (research group of the Rey Juan Carlos University devoted to Face Recognition and Computer Vision) we would like to ask for your collaboration for the acquisition of a new face database. During the first two weeks of September and in a daily basis, we will perform each of the five tests foreseen (see calendar in PDF file). We only ask for your popping by to room 153 in 2nd Department Building from 9 am to 6 pm with your passport. You will only need one minute for each test. If you experience any problem of time or calendar, we will dedicate two more days of the following week to complete the tests if necessary.
Upon completing the five tests, we will provide you with two lunch tickets (non-refundable), each on valued 5 euros, usable for the cafeterias of the Móstoles Campus of Rey Juan Carlos University.
Person in charge: José Mazorra de Cos
Where we are: 2nd Department Building, Room 153, Móstoles Campus
Timetable: 9 am to 6 pm
Poster announcing the foreseen calendar for the tests (69 kB, in Spanish)
Researchers from the FRAV group of the Rey Juan Carlos University (URJC) have developed an automatic system for driver's hands detection, providing a driving risk level which can produce a reliable answer of the different distractions of the driver. The study has been carried out in a professional lorry cabin, using a similar enviroment to the real scenario of heavy vehicles drivers.
Inserting a CD, answering the mobile phone, eating, the movement of an animal or a passanger or the presence of more and more mobile devices inside the car provoke that about 30% of the time the vehicle is moving, the drivers perform secondary tasks which can produce distractions. These deficits include failures to keep the lane or to control speed, the increase of the reaction time, missing traffic signals, dangerous mental gaps, poor attention to the road and therefore undesirable accidents.
The nature and magnitude of the deficit depends on the characteristics of the task competing with driving, so one of the main conclusions derived by the FRAV researchers is that the study of the driver's fatigue level is complicated enough for only one technique to obtain a good performance, including head tracking, eye checking or the study of the relative position of both hands. However, despite several works are necessary to complete an accurate diagnostics of the driving distractions, the study of the hands position is the least analysed so far. That is why this research performs and tests a program that allows detecting the driver's hands reliably, and measures the risk of the driving by means of an original perspective so far. Moreover thanks to this research, supervised by Enrique Cabello Pardos, we will be able to elicit a set of distraction data, above all hands-related, which will be analysed statistically.
A hardware that does not dazzle the driver:
The vision system comprises a CMOS sensor, an infrared filter, three infrared lights and two diffusers. The main component, the CMOS sensor, has a 12-bit analogical-digital conversor able to deliver more than 200 images per second in QVGA resolution (320 x 240), even achieving a maximum resolution of five megapixels. It also has a "Demo2A" card which allows a connection with a computer by means of a USB 2.0 port. An infrared filter is coupled to this sensor to block the visible light reducing the noise coming from other illumination sources, such as artificial light, and at the same time it allows using a type of lights that do not dazzle the driver as they emit non-visible light to the human eye. The system is installed on the ceiling of the cabin just above the steering wheel and the image obtained encompasses the whole steering wheel, part of the driver's legs and the gearshift, the handbrake, the GPS and some part of the copilot's seat. Such a big region allows us to determine the hands position if they are there.
The use of this vision system has showed that it provides multiple benefits with respect to other invasive ones, such as pression sensors, for instance, as it can track the hand beyond the steering wheel, if its on the GPS or on the gearshift. Another advantage of this system is about illumination, despite this is one of the most delicate parts of a vision system, as it has been shown that the use of infrared light is a solution that keeps the illumination constant enough and protects with respect to a change of illumination. As a summary, we have shown that this new technique works acceptably and in combination with other measurement systems it can produce a reliable answer of the driver's possible distractions.
Another novelty of this technique is the use of experts to define the concept of driving risk. In the definition of risk there are multiple factors to be taken into account (the state of the road, the vehicle speed, etc.), so it is not an easy problem to define. In the presented system, a group of experts define intuitively a risk signal and a program learns to predict that risk signal. These experts provide the knowledgmente that the system "learns". The results, which were checked in a simulator, show promising results.
Media that have mentioned this press release: