Deep Separable Dilated Convolutional Neural network for Cloud Detection and Removal in Satellite Imagery

The satellite is observation activities such as urban and agricultural and disaster management are constrained by adequate spatial and temporal ground observation. The detecting clouds data is usually distort ground optical observation and reduce the temporal resolution it degrades the learning valuable information. The significantly impacted by the presence of pixels affected by cloud and showdown of the cloud. This paper, proposed Lightweight Deep Separable Dilated Convolutional Neural network (DS-DCNN) technique is efficiently performed in the cloud detection removed from the satellite imagery. The data collected from the Sentinel-2 Satellite dataset and pre-processing involved the minmax normalization and uniformly scale the feature vector. Data augmentation is process to detecting various of clouds data and enhance the presences of training data to handling difficult of the satellite imagery. The DS-DCNN technique efficiently detecting clouds of varying sizes and shapes of the image from the different satellite dataset and capture global structure of satellite image and removed unwanted data. The performance of improved Deep Learning (DL) model evaluated the various metrics such as Mean Absolute Error (MAE), Mean Square Error (MSE), Root-Mean-Squared Error (RMSE) Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) metrics on Sentinel-2 Satellite dataset.