Anomaly detection in metro passenger flow is significant for the operation of metro system. Metro smart card data can provide more accurate data sources for detect anomalous metro passenger flow. This paper proposes a data-driven method based on random matrix theory (RMT) to detect anomaly in metro passenger flow. The method mainly includes three parts: matrix construction in metro passenger flow, the transform of raw matrix and the detection with RMT (M-P Law and Ring Law). Two cases are designed and conducted to validate the performance of the method based on RMT. The results indicate that the method can achieve an acceptable detection performance.