In remote sensing image processing, cloud identification in satellite remote sensing pictures is essential. Identifying between areas containing and without clouds improves the utility of remote sensing data for activities like target identification and cloud removal. In order to turn the cloud detection issue into a classification test this re-search proposes an innovative approach that utilizes a deep learning architecture. The method modifies the convolutional neural network model VGG16 as a feature extractor for training. The trained feature extractor automatically learns robust features. Finally, a random forest classifier is developed on the learnt characteristics to distinguish between thin clouds, heavy clouds, and cloudless locations. Compared with traditional feature extraction methods, the proposed approach demonstrates superior performance in classifying all three categories.