Prompt and accurate identification of anomalies in passenger flow within metro systems is crucial for safety, security, and operational efficiency. However, traditional anomaly detection methods often struggle to achieve high accuracy and low latency when constrained by limited labeled data for online applications. This study presents a straightforward yet effective online anomaly detection framework, termed multiview online passenger flow anomaly detection (MVOPFAD), to address these difficulties in a data-driven manner. Specifically, to reduce the computational burden and meet online requirements, we particularly propose an elastic extreme studentized deviate (EESD) model accounting for the characteristic of abnormal passenger flow. Concurrently, an improved shifted wavelet tree (ISWT) is employed to effectively capture various passenger flow features. It is joined by the implementation of ensemble learning techniques and EESD to further enhance the accuracy and robustness of our detection model. To evaluate the performance of our proposed framework, we conducted a comprehensive series of experiments utilizing data collected from the automated fare collection (AFC) system integrated into the Beijing Metro network. Our proposed MVOPFAD demonstrates significant superiority over the other three types of methods across all evaluation metrics. In particular, it yields a 15.49% increase in precision and a 9.71% rise in the $F2$ -score compared to the second-best model for detecting outbound passenger flow anomalies. Additionally, our model incurs lower computational cost than deep learning models and machine learning models. The experimental results strongly suggest the feasibility of online implementation, thereby demonstrating the practicality and effectiveness of our proposed model.