Driving Innovation to Fully Autonomous Cars with Machine Vision Technology

Over the past couple of years in Silicon Valley, we’ve all seen them driving around town and on our freeways – yes, those mysterious vehicles with the rooftop spinning cylinder tower mapping our roads and evaluating traffic conditions. Well, now they are becoming harder to spot as those spinning towers are getting smaller, unless there is no driver behind the wheel. Self-driving cars are closer to becoming a reality as companies rigorously test various sensing technologies to make this concept of stress-free commutes and safer roads a part of everyday life. While these cars are still in the early development stages, Samsung is actively pursuing advancements in image sensors, an important technology in autonomous driving systems, to push this ambitious vision forward.

A self-driving car has hundreds of sensors placed all over the vehicle to read, map, and react to the surrounding environment. Today, most automotive companies already use conventional image sensors, but autonomous cars need additional “eyes” in order to see further and more accurately under any conditions. In addition, these vehicles must efficiently communicate and analyze the images being captured, to facilitate speedier reactions to events such as pedestrians and other cars. In order to ensure the safety of everyone near self-driving cars, we need sensing technology with the same recognition and processing that the human eye provides, if not better.

Samsung System LSI has brought us closer to this reality through our Dynamic Vision Sensor (DVS), an innovative image sensor that mimics the human retina to instantly register changes in the nearby environment. Traditional sensors send a continuous stream of image frames to the vehicle for processing, say to understand the proximity of neighboring cars or lane dividers. But autonomous cars can require a dozen or more image sensors, and the total bandwidth of the video from these sensors surpasses conventional processing techniques, rendering the information unusable. A more efficient vision sensor is required to help bring self-driving cars closer to reality, and that is where DVS comes in.

The dynamic vision sensor, the DVS dramatically reduces data bandwidth by capturing the data that matters – that is, changes in the environment. The DVS provides an optimal balance of resolution, fast sensing speed, and low power consumption from daylight to near-dark. Through this latest innovation, we are one small but crucial step closer to a safer world where self-driving cars will exist.

If you’d like to learn more about this exciting technology, I will be presenting more details on its automotive applications at the 2017 Image Sensors Auto Americas Conference at the San Francisco Hilton on October 10, 2017 at 2:00 p.m. We hope to see you there!