Spacecraft cabin environments are complex and dynamic, posing challenges to anomaly detection and intelligent regulation. This study aims to establish an effective detection and regulation framework to address anomalies in spacecraft cabins under resource-constrained conditions. An improved YOLO-World model is adopted as the detection backbone. Two enhancements are proposed: a lightweight prompt embedding optimization using low-rank adapters (LoRA) to reduce memory usage, and a dynamic spatial attention module (DSAM) to enhance the detection of small and occluded objects. Multimodal sensor data are integrated to support decision-making.Experiments are conducted on a simulated spacecraft cabin dataset containing various anomaly scenarios. Results indicate that the proposed method achieves noticeable improvements in detection accuracy and inference efficiency compared to the baseline model.The study concludes that the enhanced YOLO-World framework can effectively support anomaly detection and environment regulation in spacecraft cabins, meeting real-time and resource limitations.