A method analysis for hail cloudy prediction based on CNN.

A hailstorm forecast method is proposed in the paper. Edge Detect Cellular Neural Network (EDCNN) method is used to extract the edge of cloud radar images. We have detected the texture of the cloud images. Then the texture image has been processed with wavelet transform. The hail data information from the image has been found. We will get approximate detail coefficients, level detail coefficients, vertical detail coefficients, diagonal detail coefficients, and reconstructed coefficients. Construct hail cloud life feature vector matrix to explain the problem. Found the corresponding rules through the five coefficients. At last, through the simulation experiment achieve the purpose of hail forecast. A feature vector of hail cloud life has been constructed, the rules of hail had been found from the feature vector. And Compared with the contour variance of the hail cloud inner and outer, we will find this paper puts forward the method is more effective. The conclusion is rationality according the simulation experiment verifies. [ABSTRACT FROM AUTHOR]