An Intelligent Typhoon Damage Prediction System from Aerial Photographs.

Damage prediction of typhoons or hurricanes is an intractable task. It needs complete meteorological information and disaster condition evaluation. It also depends heavily on the experiences of the government and evaluator. Different angles and characteristics of typhoons may cause different degrees of damage. It is a difficult task for the government to predict the possible damages correctly. This paper proposes an intelligent typhoon damage prediction system from aerial photographs. The system uses wavelet transformation, support vector machines, and fuzzy neural networks for image compression, classifi-cation, and error correction. The system then uses case-based reasoning and fuzzy damage measurement to compute possible damages to agricultural products and inhabitants. Jaiosi, Taiwan is a place used by the system as an example to illustrate its functionality. The experiment shows that the system can find important surface features from aerial photographs as well as predict possible typhoon damage correctly. [ABSTRACT FROM AUTHOR]