Cloud detection in all sky ConCam images by Gaussian fitting and valley detection in histogram

Abstract

Clouds are the complex part to classify in the all sky Continuous Camera (ConCam) image due to their intensity similar to other objects and their diverse nature. The paper presents a histogram based technique to extract clouds in ConCam images. Small scale features are extracted by subtracting Gaussian fitted histogram from original histogram of an image. Valleys are then detected in these small scale features and based on these valleys, i.e. their intensity levels, different regions of sky are classified.