Deep Learning to Take Traffic Analysis to the Next Level
SMARTCITIESWORLD (MAR 23, 2018)
Miovision developed this technology using deep learning so the system could “truly understand” the environment around the intersection
Smart city technology provider, Miovision, has introduced SmartSense technology which aims to use deep learning artificial intelligence (AI) to turn traditional road intersections into smart intersections.
According to Miovision, SmartSense brings AI to the roadside to help cities “sense and understand” what’s happening at the intersection in real time. It can detect the presence and movement of vehicles, pedestrians and cyclists and use this data to improve congestion and safety.
The new SmartSense technology completes the company’s TrafficLink solution, which also includes a 360-degree fisheye camera and an Internet of Things (IoT) connected hub that allows traffic professionals to remotely access the intersection. Together, these components make up an AI toolkit that uncovers insights about the intersection.
Like many industries, traffic and transportation are becoming increasingly data-driven. Urban planners are using traffic studies to make decisions, and safety advocates are pushing for better data sets about what’s happening on the roadways, said Miovision.
Video is the best way to collect traffic data because it is easy to install and can support a wide variety of applications. However, until now, video detection and analytics solutions weren’t good enough. Based on 15-year old technology, a small shift in the camera or adverse weather conditions could throw results out.
Miovision said it chose to develop this technology using deep learning so the system could “truly understand” the environment around the intersection, providing more robust and reliable insights.
Deep learning requires a large data set in order to properly train AI. Over the past 10 years, Miovision has accumulated the world’s largest video traffic data repository by counting more than seven billion vehicles and 770 million cyclists and pedestrians.
Miovision said this has allowed it to train its AI to recognise the movement of vehicles, cyclists and pedestrians, and extract insights such as the length of travel time during rush hour and where jaywalking happens most frequently. And it does this in real-world conditions, such as snow, rain, fog, shadows and glare.
“To describe it in human terms, the TrafficLink solution represents the eyes and brains of an intersection,” said Dave Bullock, vice president of product strategy at Miovision. “It can effectively see and understand what’s happening, respond with the appropriate actions, and use the insights gained to optimise the intersection for traffic flow and safety.”
Detecting the movement of vehicles, pedestrians and cyclists is just the first application of this technology. Miovision said its AI will enable solutions that detect and address roadside incidents such as safety hazards, crashes and double-parked cars.