Deforestation driven by anthropogenic activities poses a significant threat to ecosystems, biodiversity, and climate stability. Accurate and timely detection of deforested regions is essential for environmental monitoring and sustainable land management. This study proposes a hybrid deep learning framework that integrates MobileNet, a lightweight convolutional neural network (CNN) for feature extraction, with Random Forest, a robust classification algorithm, to enhance deforestation detection from high-resolution satellite imagery. The proposed model undergoes systematic preprocessing, including noise reduction and data augmentation, to improve image quality and classification accuracy. MobileNet extracts critical spatial features, which are subsequently classified by Random Forest, ensuring robustness against noisy and inconsistent data. The model is trained and evaluated using benchmark satellite datasets, with performance assessed through accuracy, precision, recall, and F1-score. Comparative analysis against existing deep learning models, such as ResNet and VGG, demonstrates its computational efficiency and classification effectiveness. The proposed approach offers a scalable, cost-effective solution for real-time deforestation monitoring, supporting environmental decision-making, policy formulation, and climate resilience initiatives.