DETECTION OF CLOUD SHADOWS USING DEEP CNN UTILISING SPATIAL AND SPECTRAL FEATURES OF LANDSAT IMAGERY

Abstract

Aim: The proposed work emphasizes here on detection of cloud shadows using Deep CNN (Convolutional Neural Networks) utilizing spatial and spectral features of Landsat imagery. Results: In the current study deep CNN Algorithm is used for cloud and its shadow detection. We used python libraries to create a CNN. Fourier transformation is applied on that array to transform as per their requirements. Conclusion: Using the Deep CNN algorithm, we were able to combine the whole input image to get multilevel features. Deep CNN does better image processing and semantic segmentation when compared with existing fuzzy-c and f-masking. HIGHLIGHTS: An improved approach using Deep CNN (Convolutional Neural Network) does better image processing and semantic segmentation when compared with existing fuzzy-c and f-masking.