Study On The Model Design Of Hail Forecast In Xinjiang Province

Hail is a kind of weather and caused by strong convective weather. It can cause disaster. It hindered the economic and social development of Xinjiang province and its surrounding areas. Xinjiang is located in the inland arid and desert region, it was deeply troubled by hail disaster. Shihezi is one of the main agricultural areas in Xinjiang province, and it is also plagued by hail disasters. Although the duration of hail is short, but it can caused by persistent devastating disaster. At present, the weather forecast center uses the Doppler Radar's echo series product to carry on the work to the hail. Hail this kind of strong convective weather's feature can be presented in the weather radar.This paper will be based on the Hail Radar Reflectivity image data in various regions of Xinjiang, to extract hail discrimination index, and to raise the accuracy of hail forecast. In this paper, based on the three order cellular neural network(neural networks cellular, CNN), the stability of the five order cellular neural network is proved. Five order cellular neural network and K-means clustering algorithm are used to extract the cloud edge of radar reflectivity image. The morphological characteristics of Hail Cloud Based on image recognition; Based on the first order statistical measure of the image, the two order statistical measure is used to extract the data.; The formation of hail cloud with regional characteristics based on Clustering Based on statistical variables, the hail cloud's image by K-means clustering algorithm; Regional characteristics based on the discrimination function to establish the image of hail cloud in Shihezi area by Fisher. Among them, the article on the detection of cellular neural network function, the weather radar image texture extraction method, the method of data mining and analysis of research. The article also applies to the comparative method, the classification method, the mathematical method, summarizes the method and the experiment method to verify the feasibility of the hail forecast.Paper of innovation main for, first in three order cell neural network of detection template of based Shang proposed five order template, and on five order template of stability for has proved; second, based on hail cloud has of morphology features, proposed using both inside and outside layer profile of distance sequence of variance value of size to representative the features; last, using meteorological radar reflection rate image of different dBZ value and color of relationship, extraction get a order statistics measure, and second-order statistics measure, using statistics measure value as variable, Through hierarchical clustering analysis between different areas, are differentiated on the basis of this variable, by Fisher discriminant method established in different parts of the discriminant function. Finally, established by the discriminant function training samples tested and the tests reached a higher level of accuracy.