Transitioning between Autonomous and Manual Driving
With autonomous and semi-autonomous vehicles on the roads, the booming industry of automated vehicles has been a hot topic. Tesla Motors Inc. is selling vehicles with a semi-autonomous “Autopilot” feature across the country. Uber Technologies Inc., the ride-hailing giant, has recently begun trials with driverless cars in Pittsburgh. And there are several other companies investing large amounts of money to make the autonomous vehicles a reality.
However, for the time being, the autonomous technology is mainly designed to be used on highways. The system may disengage the autopilot under various conditions such as freeway exits. The autopilot needs to manage the hand-over between the automated and the manual modes. To have a smooth hand-over, it is important to make sure that the driver is alert and ready to take control of the car before the autopilot is disengaged.
To have a smooth transition between modes of operation, Omron, a manufacturer of industrial automation controls has merged the latest camera technology, compact processors, and proprietary algorithms to introduce facial recognition technology that detects a drowsy or distracted driver. Considering the fact that one out of every six car accidents is attributed to a drowsy or distracted driver, the technology can have a huge impact even on the safety of manual driving.
The prototype, presented at CEATEC, employs an infrared camera to monitor the driver’s eye movements and gestures. The infrared option is there to help the system tolerate difficult lighting conditions, especially at night. When the driver’s blinking deviates from the normal pattern or the fatigue causes his/her head to bob, the system wakes them up by triggering a warning buzzer. Vibration of the steering wheel and/or the seat can also be an alternative alert in the future. When the system detects that the driver is not alert and gets no reaction from the driver, it slows down and stops the car automatically.
To constantly monitor the driver’s vital signs, Omron’s technology incorporates a pulse reader and a blood pressure gauge. These devices, which will be embedded in a wrist-worn wearable in the future, attempt to detect drowsiness or an imminent stroke.
Facial recognition can detect a distracted driver. Image courtesy of the Embedded Vision Alliance.
Years ago, image processing was restricted to applications like factory automation mainly due to the expensive, power-hungry processors required. However, now we have powerful processors, image sensors, and robust algorithms to achieve practical computer vision with an in-cockpit camera.
Considering the rapidly growing market of Advanced Driver Assistant Systems (ADAS), Strategy Analytics has estimated that, by 2021, each year, more than $25 billion will be spent on the ADAS products in which image capture and vision processing will be a key element.
The estimated investment on ADAS. Image courtesy of the Embedded Vision Alliance.
Other Potential Applications of Face Recognition
Facial recognition can potentially increase the safety in car accidents by a second mechanism, as well: it can adjust the intensity and location of the airbags based on the driver’s head and body position.
Although the main application of the intelligence achieved by the in-vehicle camera is increased safety, many other aspects of driving—like comfort and entertainment—can be improved too. For example, based on who is driving the car, the system can tune the radio to the driver’s favorite stations or it can customize the navigation system by preloading desired locations. Many settings such as seat, mirror, and steering wheel options can be automatically adjusted as the driver is recognized.
The system would be able to detect if kids are in the car and, if so, restrict entertainment systems to more family-friendly content. Such a smart system can adjust the air conditioning based on the number and location of the occupants.
Facial recognition can also be used as an alternative for a login password. As a result, it would be possible to impose certain restrictions on some of the designated family members driving the car—e.g., teenagers with a new license.
"Road rage" prevention is another application that has been investigated by researchers at Switzerland's École Polytechnique Fédérale de Lausanne.
Omron’s Previously Reported Technology
Although not much technical detail was given about Omron’s prototype presented in CEATEC, the company’s portfolio has other facial recognition boards such as the B5T HVC.
The complex proprietary algorithm utilized in this board can even guess the emotions on a detected face. The board, which measures just 60 x 40 mm and consumes less than 250 mA, and can also estimate the gaze direction that is helpful for telling if the driver is looking at something other than the road. The gaze direction estimation is already used in commercial cameras to detect if a subject is looking at the camera or not.
The facial recognition product can output a lot of useful information such as the vertical and horizontal gaze direction, blink degree for each eye, and the captured image itself.
B5T has many functions such as estimating the age, gender, and more. Image courtesy of Omron.
Facial recognition systems must tolerate difficult lighting conditions and situations in which something, such as a hat or glasses, prevents a clear view of the driver. Otherwise, the driver may decide to turn the system off due to annoying false alarms.
Facial recognition technologies may use several 2D cameras or a single 3D camera that can discern depth. To process the cameras' readings, the engine control unit (ECU) or other available resources can be used.
Obviously, the number of the false alarms can be reduced significantly by employing several cameras instead of one.
Omron is collaborating with several domestic and overseas manufacturers to commercialize the new technology by around 2020.
With the growing interest in autonomous vehicles, Omron’s technology and similar inventions will play a critical role in providing a safe ride.