The Application of Ant Colony Algorithm in Meteorological Satellite Cloud Pictures Segmentation
|School||Harbin Engineering University|
|Course||Navigation,Guidance and Control|
|Keywords||meteorological satellite cloud pictures typhoon mathematical morphology ant colony clustering algorithm GLCM|
Typhoon (tropical cyclone) is one of the worst natural disasters that seriously influence human being. Strong winds, heavy rain, storm surge and other disasters caused by Typhoon have brought great inconvenience and danger to people’s production and life nearby southeast of sea in our country, even seriously threaten to the sailing security. In order to reduce disasters effectively caused by typhoon, reinforcing the real-time monitoring, location and forecasting for the typhoon is particularly important. Due to accuracy of the typhoon centre locating mainly depends on the segmentation effect of Typhoon. Therefore, the typhoon segmentation was highly valuable to study and extremely important practical significance.Ant colony algorithm is a new simulated evolutionary algorithm. Its characters of discreteness, parallelism, positive feedback and robustness make it get widely application in the image segmentation field. This paper selects FY-2 meteorological satellite cloud pictures as subject investigated,13th typhoon—TALIM as an example which landed PingHai town, PuTian city, FuJian province at 14:30 of September 1st 2005. Ant colony algorithm and its improvement are applied to the research of the typhoon segmentation, and this paper mainly involves the following four aspects of works:Firstly the research progress of ant colony algorithm and image segmentation in recent years is summarized. The principle and mathematical model of ant colony algorithm is deeply analyzed and elucidated. The realization process of ant colony algorithm is explained by solving the TSP problem. Finally, some improved algorithms of ant colony algorithm widely used in the image segmentation field are quoted.Secondly the related concepts and mathematical description of cluster analysis is briefly illuminated, and then some representative ant colony clustering algorithms in recent years are summarized. After that, four basic models are introduced briefly and merits and drawbacks of ACCA are compared. Ant colony clustering algorithm which based on the principle of foraging is used to segment the typhoon image.According to the discrete characteristics of digital images, the ant colony clustering algorithm is applied to the segmentation of the typhoon from view of the clustering. By introduced the initial clustering centers and guidance function, the large calculation and the search time of traditional ant colony clustering algorithm can be shorten obviously. The problem of low accuracy may be caused by the uneven distribution of unrelated clouds which is similar to typhoon when the ant colony algorithm is utilized respectively. To solve the problem, an ant colony algorithm integrated with mathematical morphology of typhoon segmentation is presented.Considering the typhoon image is a natural texture image, based on the grayscale characteristics, using the high-level image features--texture features can describe fully the image information for better segmentation results. Texture features of typhoon image are extracted by using GLCM and the feature vector formed.Then the typhoon-cloudy is separated from motley cloudy by using ant colony clustering algorithm.By using MATLAB7.1 as a programming tool, the effectively and accuracy of algorithm has to be tested and verified. Experiments also show that the algorithm of this article is effective to segment the typhoon-cloudy and extract the feature of typhoon area, and finally realize the location of typhoon center. The result of its research has certain practical significance for identification, forecasting and location of typhoon.