Typhoon center positioning method based on deep reinforcement learning

Abstract

The invention discloses a typhoon center positioning method based on deep reinforcement learning. According to the method, the perception ability of deep learning and the decision ability of reinforcement learning are combined, a typhoon center positioning problem is converted into a series of decision behaviors of searching a typhoon center on a satellite cloud picture by utilizing a search box,a Markov decision-making process is used for modeling in the search process, an intelligent agent is trained through a deep reinforcement learning algorithm to learn and move and reduce a search box through simple operation, the center of the search box is made to be close to a real typhoon center constantly, and then typhoon center autonomous positioning is achieved. According to the method, typhoon detection and center positioning of different levels and different forms are achieved, and the effectiveness of the method is verified through experiments.

Description

technical field [0001] The invention relates to the fields of meteorology and machine learning, in particular to a typhoon center positioning method based on deep reinforcement learning. Background technique [0002] A typhoon is a deep low-pressure vortex cloud system that mostly occurs over tropical and subtropical oceans. Typhoon is characterized by strong suddenness and great destructive power. It not only has a huge impact and disaster on aviation and navigation activities, but also causes major economic losses and casualties on the land where people gather after landing. [1] . Typhoon center positioning is an important link and key technology for analyzing and forecasting typhoons and reducing typhoon disasters. Its positioning accuracy is closely related to the determination of typhoon track, the prediction of typhoon track, and the forecast quality of thunderstorms and strong winds. Therefore, the typhoon center positioning technology The research of typhoon is the...

AI Technical Summary

Problems solved by technology

[0005] At present, these methods can only deal with typhoon cloud systems with obvious typhoon eyes or cloud bands with prominent helicity. Such typhoon cloud systems usually only occupy a period of maturity in the complete typhoon life cycle. Therefore, it is necessary to complete typhoon positioning and Typhoon track fitting is relatively difficult; and for complex typhoon image problems that are difficult to abstract representative mechanism features, existing methods are even more difficult

Images