對多普勒雷達圖像中的風暴進行準確的識別和跟蹤，是下一步進行天氣預報的前提。目前雷達圖像的主要跟蹤方法是進行相鄰時刻圖像中風暴的匹配，這些匹配方法會產生一定的誤差，特別是在風暴分裂與合并的時候。基于這一問題，將一種改進的level set方法用于雷達圖像中風暴的跟蹤。首先在第一幅雷達圖像中識別出風暴的初始輪廓，利用改進的算法，即不斷交換輪廓內鏈和外鏈的格點，將初始輪廓不斷演化確定出第二幅圖像中風暴的輪廓，依此再跟蹤下一時刻的風暴。由于輪廓的演化是基于格點的，所以在風暴發生分裂或合并時，可以實現輪廓的準確跟蹤，進而為災害性天氣的預警工作打下良好的基礎。

Abstract:

?The accurate identification and tracking of storms based on Doppler radar images is the premise of weather forecast. At present， the main tracking method of radar image is matching storms between adjacent images， which can cause some errors， especially at the time of storms splitting and merging. Based on this problem， an improved level set method is used to track storms in the radar image. Initial contours of storms in the first radar image were identified， and contours of storms in the second image were the evolution of initial contours， which was achieved by the improved level set method， i.e. switching elements between the inside lists and outside lists of contours， and so were the next radar images. As the evolution of contours is based on the grid， the accurate tracking of contours can be achieved at the time of storms splitting and merging， and finally this improved method can build a good foundation for the early warning of severe weather.

?

?

?