Automatic Cloud Detection and Weather Forecasting Using Gradient Based SPS

Abstract

Thin cloud detection is challenging in a ground-based sky-imaging systems due to low contrast and vague boundaries between cloud and sky regions. In our proposed algorithm, a series of super pixels could be obtained adaptively by SPS algorithm according to the features of clouds. In addition to that a gradient feature is drawn out from that super pixels for improving the detection of cloud types . due to its high performance in solving complex issues, simplicity of execution and low computational complexity. The nearest neighbor classifier k used . High thin clouds, high patched cumuliform clouds, stratocumulus clouds, low cumuliform clouds, thick clouds, strati form clouds and clear sky are the seven different sky condition are distinguished. Firstly local threshold for each super pixel is calculated and then determining a gradient values for the whole image. Finally comparing with the obtained threshold matrix can detect the cloud. Experiments on real natural images are conducted to show the performance of the proposed superpixel segmentation algorithm.