AI-based drone by KARI/Sungkyunkwan University wins in world contest for object detection
-
Department
Administrator
-
Registration Date
2020-10-08
-
Hits
5061
AI-based drone by KARI/Sungkyunkwan University wins in world contest for object detection
- Proven the excellence of drone image-based object detection technology -
- To be applied to the swarming reconnaissance drone system for use in search for missing persons -
□ DroneEye2020, the joint team of the Korea Aerospace Research Institute (KARI) and Sungkyunkwan University, won first place in the object detection category at the “World VisDrone Competition 2020.”
ㅇ The VisDrone Competition is an annual event held by the I(E)CCV (International / European Conference on Computer Vision), the world’s most renowned academic society in AI, since 2018. The competition evaluates the excellence of technology that detects objects, such as people and cars, using AI from images captured by drones.
□ This competition was participated in by 36 teams, including the University of North Carolina in the United States, Fraunhofer Institute in Germany, and University of Science and Technology of China. The KARI/Sungkyunkwan University team won first place for the technology of recognizing the sizes and shapes of objects, -- which change according to the drone’s flight altitude -- such as people and cars on the ground.
ㅇ The technology for recognizing frontal objects from the vehicle height has been developed for self-driving vehicles. Note, however, that the drone image-based object detection technology developed by KARI detects objects by learning to recognize objects on the ground considering the object size and shape that change according to the drone’s flight altitude and view.
□ KARI plans to install the multi-object detection drone technology on the swarming reconnaissance drone system to be used for searching for missing persons in a wide area. It can also be used for monitoring residual fire and military reconnaissance by replacing the mission sensor. In addition, it can be utilized to identify the location and number of people or cars at intersections or downtown squares with drones.