Existing deep learning-based hyperspectral anomaly detection methods often overlook frequency domain features, hindering the ability to effectively distinguish between background and anomalies.