Wind Field Retrieval For Typhoon Based On RVM And Wavelet Transfrom

It is a global problem how to reduce the harm caused by the typhoon around the world. China is one of the countries affected by the typhoon disasters, every summer and autumn, typhoons frequently struck in Guangdong, Taiwan, Hainan, Fujian and other areas. Not only makes these coastal suffer serious losses, even inland areas are affected, poses a serious threat to the people’s production and life. At present, we solve this problem mainly through a variety of methods of the typhoon forecast. As the typhoon wind field (Especially the Typhoon kernel) contains the typhoon’s intensity and path information. The high quality of the typhoon wind field can improve the accuracy of the numerical typhoon forecasts.This paper selects typhoon cloud images from2005to2008as research objects. Relevance vector machine (RVM) and wavelet transform are used to study the typhoon wind field inversion. Mainly contains three parts as follows:(1) Establish typhoon wind field characteristics model for eyed typhoon and no eye typhoon based on RVM;(2) Estimate the speed of typhoon wind field points, and access to typhoon wind field of two-dimensional scalar;(3) Analysis the direction of the motion in typhoon wind field, and get the vector of typhoon wind field based on Discrete stationary wavelet transform and Satellite images. Specific processes are as follows:(1) Establish typhoon wind field characteristics model for eyed typhoon and no eye typhoon based on RVM.The typhoon wind field model is established through two sets of data, which are the size of the typhoon eye and the radius of maximum wind. A mature typhoon cloud image may contain typhoon eye or not. If the typhoon cloud image contains typhoon eye, we can directly pre-process the typhoon cloud image, such as image enhancement and image segmentation, in order to extract the typhoon eye. While the typhoon cloud image does not contain typhoon eye, we need to artificial construct a typhoon eye, and then do further study like eye typhoon. On the typhoon cloud images, the radius of maximum wind speed is the distance between the warmest point of eye wall and the coldest point of cloud. With the data of typhoon eye size and radius of maximum wind, we can establish the mathematical relationship between the two sets of data based on the relevance vector machine, to describe the typhoon wind field characteristics.(2) Estimate the speed of typhoon wind field points, and access to typhoon wind field of two-dimensional scalar.As the gray information and the wind field information of typhoon cloud images are inextricably linked, we need to use the gray information and maximum wind speed data to establish a model based on the relevance vector machine. Then estimate the typhoon wind speed by using linear interpolation. Comparing with other common fitting methods, the relevance vector machine performs better in the running time and accuracy.(3) Analysis the direction of the motion in typhoon wind field, and get the vector of typhoon wind field based on discrete stationary wavelet transform and satellite images.In this paper, the motion direction in typhoon wind field was analyzed by image matching method in discrete stationary wavelet domain. Typhoon wind field motion vector was obtained and could be used to descript typhoon cloud evolution.Traditional gray-based image matching algorithm is simple and direct, but the large amount of computation reduces the speed of operation. In this paper, we combined the discrete stationary wavelet decomposition technique and the gray-scale image matching algorithm, and proposed a level matching method for typhoon cloud images in wavelet domain. As this method reduced the search space, the computing speed was improved, and finally achieved good match results.