Multi-stage recognition scheme for urban road construction intrusion using fiber-optic distributed acoustic sensing

Abstract

Urban road construction threatens buried utility networks, causing significant economic and safety risks. Traditional monitoring scheme faces high costs, limited coverage, and slow response times. This paper introduces an urban-scale multi-stage recognition scheme using fiber-optic distributed acoustic sensing (DAS) for real-time road construction intrusion monitoring across existing urban telecom networks. The multi-stage approach includes: prejudgment, efficiently filtering potential intrusion regions; recognition, using a hybrid deep learning model for event classification; and review, enhancing reliability through a spatial-temporal continuity mechanism. A comprehensive urban intrusion dataset was created featuring distinct vibration patterns from seven interference and eight construction events. A challenging 14-day field test across 60 km of urban environment validated the approach, achieving 99.25 % localization accuracy, 98.80 % classification accuracy, and a response time of 0.99 s. This scalable, cost-effective solution for infrastructure protection integrates with existing telecom networks and offers potential applications in urban security and emergency response.