Classification of Cloud/sky Images based on kNN and Modified Genetic Algorithm

The main objective of this paper is to identify the types of clouds from the high-dimensional feature set extracted from sky/cloud images. The identification of types of clouds finds its scope in prediction of weather, natural disasters like storms, thunder storms etc. The feature set of sky/cloud images are fuzzy, and incomplete and highly spatial and temporal. Hence, it is necessary to optimize the feature set to for further analysis. Neighborhood Component Analysis (NCA) is used for optimizing the feature set extracted from sky/cloud images and it also improves the speed of classification. K-Nearest Neighbor (kNN) classifier is combined with Genetic Algorithm that uses Tabu search is used for classification. Finally, this paper identifies the following types of clouds. (i) Pattered cloud (ii) Thick dark cloud (iii) Thick white cloud and (iv)Veil cloud. kNN algorithm is modified to include condensed training set representation to improve the classification performance. The experimental results are verified with Singapore Whole sky imaging categories (SWIMCAT) database and achieves better performance in classifying the types of clouds.