Satellite Image Cloud Classification Using Fuzzy Logic

Abstract

Clouds are important to our planet. They affect the earth's energy and balance climate and weather, depending on the cloud type. Information from clouds taken from remote sensing presents challenges; time is needed for reprocessing, and the texture of a cloud is changed under atmospheric conditions. Obtaining accurate and automatic cloud classifications is still the main problem. In this research we build a system that classifies satellite cloud images into three types: high, mid, and low cloud images. The approach that is used to build the system is search and acquisition of satellite cloud images from a website. 120 satellite cloud images were collected. Then the images were reprocessed by converting them into gray scale images, applying median filter and resizing the images into 100×100. The feature extraction used on images was histogram of gradient. The result was 4356 feature vector. Using the fuzzy logic inference system, the rules of the system were depended on the correlation between the cloud type and the mean (average) of feature vector of the histogram of gradient for each cloud image. The Trapezoidal shaped membership function is used for input value and Triangular shaped membership function is used for the output value. The result of FIS was 97 % accuracy with 0.25 second.