IMAGE SEGMENTATION OF TYPHOON SPIRAL CLOUD BANDS BASED ON SUPPORT VECTOR MACHINE

Typhoon is one type of disaster weather which can impose serious impact on the life and production of human society. It has special physical characteristics of clouds with the structure of cloud eye, cloud walls and spiral cloud bands. Much information about Typhoon motion, wind field and heavy rainfall is contained in spiral cloud bands. Therefore it is very important to segment and recognize its spiral cloud bands as clearly as possible. As it is difficult to achieve the expected result with the existing image segmentation algorithms, in the study support vector machine was applied to segmentation of Typhoon spiral cloud bands while the problem of image segmentation can be solved by transforming it to the problem of classification. In order to achieve ideal results of segmentation, it is necessary to analyze and compare all aspects carefully in the selection of training samples, the determination of kernel function, the setting of its parameters and the common segmentation algorithms. The results demonstrated that the algorithm proposed in this paper has much more advantages than the existing algorithms, being more accurate, rapid and robust.