颱風在台灣是一種常見的天災。颱風帶來的充沛雨量讓台灣不受乾旱的威脅,但是一個威力強大的颱風往往在台灣各地造成巨大的災情,因此颱風預測便會顯得相當重要。颱風眼與雲層溫度變化的差異在颱風識別中是很重要的特徵,在本論文中,我們藉由分析紅外線衛星雲圖的方式,來找出颱風雲層在衛星影像上的特徵。許多研究都是利用切割過後的颱風雲層做颱風影像的定位,所以在衛星雲圖中,雲層的分離與切割是相當重要的一環。透過幾個前處理的方式,我們對颱風影像中的雲層做切割,同時,我們提出對所有切割雲層影像的像素作切片分層的分析,藉以觀察雲層溫度變化的範圍,最後找出颱風的雲層,並且透過颱風眼與颱風中心周圍像素值的差異來達到定位的效果。
Typhoon is a common nature disaster in Taiwan. However, typhoons also bring plenty rainfall so people in Taiwan do not have too worry about the drought. A strong typhoon will always make a serious destruction in its path. Therefore precise typhoon prediction is very important. The eye of the typhoon and the variation of the temperature change of cloud are important features in the typhoon identification. In this thesis, we analyze the infrared satellite cloud images and extract the features of typhoon in the satellite images. Many researchers used the segmented typhoon cloud images to locate the center of the typhoon. So separation and segmentation of the cloud in the satellite image are important topics in this thesis. Several preprocessing steps are done to facilitate the segmentation of the cloud from the typhoon images. Meanwhile, we present methods that analyzed all the segmented cloud slices based on pixel value and observe the cloud temperature change. Finally, we find the typhoon cloud and locate the center of typhoon according to the difference between the eye of the typhoon and the cloud around it.