In this paper we discuss on a system where in automatically the images collected from the spacecraft are identified for the connected images. the satellite images taken consist of not only basic shapes with regular size and shape but also the interconnected irregular shapes which is identified as network. Now the images that are automatically classified are either normal weather patterns or adverse weather patterns developed. An early warning is raised based on the images classification if the tendency of adverse weather system is about to be formed. In order to do this the strong knowledge of angular movement of the clouds, the knowledge about clouds, image processing and Neural Networks techniques are essential. The knowledge about clouds helps in analyzing the image data based on which decision can be taken. The image processing knowledge helps in classifying the images and the Neural Networks techniques helps in processing the interconnected images of the selected region.
Since the cloud formation is arbitrary in nature, identification of clouds as segments in the image has to use the techniques of Neural Networks.