Computer Vision in Automotive

Dear Friends,

Artificial Intelligence can have a huge impact on how human civilization moves forward. In this blog, I am excited to talk about how advances in Computer Vision will impact automotive.

I am a big fan of Science Fiction and the Star Trek series. The prospect of an omnipresent, all-seeing computer is both exciting and terrifying. How will it be to travel in a car that navigates itself and takes care of passenger safety and comfort automatically?

Advanced Driver Assistance Systems (ADAS) and Autonomous Cars

Cars with the ability to alert the driver when there’s an inadvertent lane change or when the driver gets drowsy are poised to become mainstream. Some of the other ADAS technologies that are coming up are Traffic Sign Detection, Pedestrian Detection, Collision Warning, Blind Spot Detection etc. These ADAS technologies use a combination of Radar, Lidar, cameras and other sensors.

Then comes the prospect of the fully autonomous self-driving cars. Google has been testing one in California and Nevada for a while now. Recently there was the news report of Delphi’s self-driving car doing a coast to coast drive in the US. Similar to ADAS, self-driving cars also use a combination of Radar, Lidar and cameras to enable the car to navigate on their own.

In today’s ADAS systems and self-driving car prototypes, Computer Vision technology using multiple cameras plays a key role in many applications. These include Lane Departure Warning, Traffic Sign Detection and Driver Drowsiness Detection. Bird’s Eye View is a camera based technology that is already getting deployed in many markets. Some of the other requirements such as Collision warning or pedestrian detection can be implemented using both cameras or Radar or Lidar.

As the computer vision technology matures and is able to match the accuracy of Radar and Lidar, more and more functions will be implemented using cameras. This is because cameras are much cheaper than Radar or Lidar.

Driver awareness

As cars become more autonomous, it will become increasingly important for the car to be aware of the driver’s status. Is the driver drowsy, sleeping or is he alert to take control in the event of an emergency? In addition, as the role of the driver diminishes in driving the car, the driver will expect the car to be more comfortable. Personalized entertainment or seating comfort will start becoming more relevant.

Understanding the status of the driver will require an inward facing camera backed by advanced Computer Vision algorithms. This camera can also be used for recognizing the occupants of the car and providing personalized comfort and entertainment. Camera based eye and gesture tracking for radio HMI and cockpit control are being developed by some of the leading car makers on an experimental basis.

What is AllGo’s play in this space?

AllGo is leveraging its field proven AllGoVision video analytics technology to develop Computer Vision algorithms that are relevant for the automotive. Driver drowsiness detection and face recognition based personalization are currently available, and other CV algorithms for automotive are in the works. With an eye on autonomous cars of the future, AllGo will combine our Infotainment and Computer Vision solutions to bring unique innovative solutions to the market.

What does the future hold?

I will stick my neck out and predict that mature Computer Vision technology will be imperative to make the self-driving car fully autonomous. A recent Consumer Watchdog report points out that the self-driving cars cannot understand a human hand signal or a traffic cop with hand signalling, and thereby cannot be fully autonomous. It is clear that such problems can be overcome only by Computer Vision – by giving the car the ability to understand its surroundings using cameras, similar how our brain understands using our eyes.

Can I watch the world cup soccer match live on my way to work while the car reaches me there safely? Will the car be able to drive through crowded roads of a city, navigating the traffic cop showing the hand signal or the pedestrians who walk across unexpectedly, I wonder!

Only time will tell! The next decade promises to be exciting for Computer Vision in automotive.