Anomaly Intrusion Detection Algorithm Based on Distributed Fiber Optic Sensors

In the field of cultural heritage preservation, distributed fiber optic sensors are widely adopted for their long transmission distances, minimal loss, and suitability for long-distance monitoring. They are frequently used to monitor large perimeter intrusion events. Traditional intrusion detection methods are typically employed in fixed scenes without considering weather conditions. In volatile situations, these methods often exhibit lower recognition rates for human and mechanical disturbances and higher false alarm rates. To address these challenges, this paper proposes an anomaly intrusion detection algorithm, CCL-DCT-ST, integrating weather information. The main contributions are as follows: considering weather information in experimental monitoring scenarios, proposing a CCL detection model, and integrating a DCT-ST dynamic confidence threshold self-training strategy to identify abnormal intrusions effectively. Experimental results demonstrate a 25% improvement in accuracy on the dataset used in this study with the effective utilization of weather information. Moreover, implementing the DCT-ST semi-supervised strategy shows significant accuracy improvements on unlabeled datasets across various scenarios, thereby reducing false alarms from distributed fiber optic sensors.”