A fast cloud detection approach by fuzzy C-mean and adaptive threshold on satellite imagery

Change detection allows direct observation of land surface at repetitive intervals and provides important applications in environment monitoring, damage assessment and so on. However, cloud detection is a precondition for deriving land change information of satellite imagery of different dates. To have high operating efficiency, we proposed a fast cloud detection approach based on Fuzzy C-Mean and adaptive threshold before detecting a changed area. It is a two-step algorithm, firstly using Fuzzy C-Mean clustering to dilute the small brightness area even with noise and outliers. Then setting a different threshold at each band to find the cloud area. Experiments were run on several images acquired under different conditions, the preliminary results indicate that the proposed method significantly improves the performance of cloud detection.