Satellite images are very important information for the control of disasters, field planning, and monitoring of the environment. However, the usage of satellites to obtain image data introduces high dimensionality, which is an issue in terms of processing time. The approach for feature selection that this paper presents incorporates a Genetic Algorithm for selecting the best features in improving classification performance with deep learning feature extraction. The proposed method of segmentation sorts images into classes including cloud, desert, green area, and water area. The deep features are extracted with the help of the MobileNetV2 architecture that is pre-trained on the ImageNet dataset, the benefit of which is in acquiring multi-layered representations of spatial patterns and semantics from the satellite imagery. Next, a genetic algorithm is used for selecting the best subset of features from the huge feature space for better classification with high accuracy. It applies a purely genetic strategy in its iterative evolution of a population of feature subsets selected according to the discrimination level as provided by the SVM classifier. The ideas in this integrated framework effectively diminish the feature dimensionality and, in some cases, augment the classification performance in comparison with the conventional methods. The efficiency of the proposed model is determined with the help of performance indicators including accuracy, precision, recall, and F1-score. The obtained results, therefore, imply possible benefits from implementing the integration of evolutionary algorithms for choosing feature subsets and deep learning for extracting features in satellite image classification systems.