In planetary exploration, crater detection is vital for robotic navigation, geological history research, and terrain analysis. This study enhances crater detection accuracy by modifying the YOLOv9 deep learning model through the addition of Enhanced Channel-Aware Attention (ECAA) layers, stochastic transformers, a Precision-Optimized Feature Pyramid Network (FPN), and a custom loss function. To enable adaptation to different planetary surfaces, the model was initially trained on a publicly available lunar crater dataset before being fine-tuned on a combined dataset including craters from the Moon, Mars, and Mercury. Results show that the modified YOLOv9 outperforms standard models, detecting craters of varying sizes and visibility with improved precision and recall. This approach provides an efficient and scalable method for planetary mapping and space exploration by enhancing crater detection accuracy across diverse celestial bodies.