This paper introduces a real-time decision-making model for drilling site selection, leveraging three-dimensional convolutional neural networks (3D CNNs). The model is devised to mitigate multi-source errors and enhance the success rate of drilling operations. Through the generation of local three-dimensional geological structure maps, the model achieves a precise depiction of the geological structure within the drilling area. Moreover, it integrates various factors such as terrain slope and rock distribution to formulate the drilling site selection decision-making model. Experimental validation demonstrates that the model exhibits high accuracy and real-time efficiency in lunar soil excavation tasks. The research not only provides robust support for lunar exploration endeavors but also furnishes valuable insights for optimal decision-making in deep space exploration and resource utilization domains.