A Fine-Tuned CNN Framework for Satellite Image Classification and Area Estimation

Satellite images analysis is required in land-use planning, environmental monitoring and agricultural evaluation. This paper presents a concisedeep learning architecture that consumes state of the artconvolutional neural network (CNNs) to estimate fieldareas and classify satellite images. To classify ssatelliteimages into four classes (Cloudy, Desert, Vegetation,and Water), the proposed approach will involve imagepreprocessing, augmentation, and an optimized CNNmodel. Field area is estimated by pixel-wise analysis ofthe outputs of a classification. The procedure is suitablein the real-life setting of remote sensing applicationsbecause experimental findings provide high level ofclassification and reliable determination of areas.