Method Research Of Typhoon Cloud Recognition And Its Center Location Based On Doppler Weather Radar Image

Typhoon, as one of the disastrous weather, imposes serious impacts on socialproduction and human lives. Therefore, it is practically significant to take exactsegmentation and locate accurately the typhoon center on the basis of Dopplerweather radar reflectivity images. The paper has proposed a series of methods inhelping the segmentation and recognition of typhoon cloud bands, regionalmodification and abstraction of skeleton lines as well as the location of the typhooncentre.It can be considered as the purpose to separate the spiral cloud bands fromTyphoon cloud blocks. The segmentation of typhoon is targeted at acquiringinformation of typhoon cloud bands within Typhoon could blocks. This paper hasproposed a segmentation approach based on Support Vector Machine. With theselectionofstableordynamicsamples,awell-trainedimageclassifiercanbeobtainedon the construction of SVM. As a result, it is efficient to distinguish the spiral cloudbandsfromtheDopplerweatherradarreflectivityimages.After the segmentation of spiral cloud bands, it is necessary to distinguish thetypical spiral cloud bands from those cloud blocks which don't have the distinctcharacteristics. In order to solve the problem, this paper has proposed a regionrecognition approach based on feature description and classification discriminance.With the feature extraction and the training for SVM, a classifier can be acquired,which results in a clear separation of the typical spiral cloud bands from the otherTyphooncloudblocks.The acquired typical spiral cloud bands need the regional modification with itsskeleton lines to be shaped and smoothed. This paper has proposed an effective wayusing mathematical morphology to smooth the edges of them. On the basis ofmodification and smooth operation, intact skeleton lines can be achieved with veryfewbranchesbyskeletonlineextractedalgorithmandpruningalgorithm.It is completely dependent on the acquired skeleton lines to position centerlocation of Typhoon. This paper has proposed a method using genetic algorithm andtemplate match to locate the most likely central point. Through optimizing thetemplate parameters, the best matching result will be gained and thus the most appropriatecenterlocationwillbeconfirmed.In this paper, the problems of segmentation of spiral cloud bands and location oftyphoon centre are solved with the application of pattern classification andevolutionary computation on the basis of the Doppler weather radar reflectivityimagesandthecharacteristicsofdevelopingTyphoon.