This research presents an intelligent framework for the automatic classification of ground targets in Synthetic Aperture Radar images. The principal challenges addressed are the diversity of target characteristics and the presence of inherent speckle noise, which complicate manual analysis and classification. To address these issues, a hybrid architecture based on a Convolutional Neural Network and a Genetic Algorithm is proposed. In this method, the Genetic Algorithm is employed for the automated optimization of the CNN architecture's hyperparameters, while the Lee filter is utilized in the pre-processing stage to effectively reduce noise and preserve edges. Performance evaluation on the standard MSTAR dataset demonstrates that the proposed method achieves a remarkable accuracy of 99.33% in classifying the denoised images. This result clearly indicates the superior performance of the proposed method compared to other conventional approaches. The outcome of this research confirms the high efficacy of the hybrid approach in addressing the complex challenges of SAR image processing and represents a significant step towards the development of robust Automatic Target Recognition systems.