An anomaly detection method based on feature regular constraints

Abstract:

Atmospheric jet is one of the important factors affecting weather patterns, and the jet identification can provide important data for automatic identification of weather system and intelligent weather forecast calculation. According to the property that jet axis locates at the geometric center of strong wind area，a new jet axis identification method is presented. First of all, the coherence feature of wind direction is used to obtain the jet areas, and skeleton extraction method is employed to get area geometric skeleton. The jet axis is obtained by connecting the tuned skeleton points which is computed by gravity formula. The experiment basing on MICAPS 11th format data and comparing with manual way and other methods shows that the present method can get the jet axis effectively.