Convective objects detection using computer vision based on radar images

Weather radar is a meteorological observation equipment that can detect meso-scale meteorological phenomena as well as small-scale convective phenomena at an early stage with high temporal and spatial resolution. This study intends to introduce radar data used in meteorology into other perspectives. Although the physical meaning of radar data variables is more generally used, actual weather forecasting agencies are more sensitive to visible changes in the radar image itself. Computer vision technology is a representative way to directly utilize radar images. It produces new information by identifying and interpreting objects or situations by imitating the human visual cognitive system. In this study, convective echo changes were detected and tracked by computer vision using high temporal resolution of X-band weather radar images observed at 1-minute intervals. For convective echo detection, the background subtraction method was used, and techniques such as the median value and Gaussian filter were also used to minimize the echo detection error. The paths of the detected echoes were predicted using the Kalman filter technique. In order to classify the detected multiple convective echoes into individual objects, the labeling of individual convective echoes was also implemented by giving a certain size and intensity as conditions. Finding and tracking convective echoes in real-time with high spatio-temporal resolution radar images will help detect signals early that can generate severe weather and respond preemptively to hazardous weather.

 

Acknowledgements:

This research was supportedby the "Development of analysis technologies for local-scale weather radar network and next generation radar (KMA2021-03221)" of "Development of integrated application technology for Korea weather radar" project funded by the Weather Radar Center, Korea Meteorological Administration.