Electronics and Telecommunications Research Institute (ETRI), South Korea's research institution, announced on October 26 that it has won the first place in the "object segmentation and tracking technology for autonomous driving" at the International Conference on Computer Vision (ICCV) held for 6 days from October 11th.
The competition is a competition in which several objects are divided and tracked based on road images taken from the point of view of an autonomous vehicle.
With an algorithm developed through international joint research, ETRI analyzed the video provided by the organizer and tracked about 20 objects such as roads, walls, traffic lights, buildings, and people.
The research team's technology divides objects into pixel units to recognize shapes and color them. Therefore, detailed identification of objects and precise tracking are possible.
It is a much more advanced technology than the existing method of recognizing and tracking objects with a rectangular frame.
This algorithm determines whether or not each pixel is an object by itself, and includes a technology for more accurately tracking the change in the position of the object.
ETRI said that it was able to achieve the best record in the competition by using the contrast learning technique to more accurately recognize the associations between objects.
The research team revealed that this technology is specialized in object segmentation and tracking for autonomous vehicles.
The team explained that they confirmed that the performance was superior to other technologies in various environments such as weather, lighting changes, object size, occlusion, and street environment.
Object segmentation and tracking technology is a technology that can accurately and quickly recognize the location of vehicles and pedestrians on an intersection or road.
When applied to traffic control systems for smart cities in the future, it is possible to increase safety and link various services.
For example, it is possible to send a warning signal to both pedestrians and drivers using the crosswalk by accurately figuring out and predicting the direction of vehicles entering and exiting the intersection. This can dramatically improve the risk of traffic accidents.