SMART-TOLL: A Behavioral Frequency and Spatial Pattern Based AI Framework for Automated Residency Verification at Highway Toll Plazas

Highway toll plazas in India face a persistent challenge: accurately identifying local residents eligible for toll exemptions versus tourists who must pay. Current manual verification is slow, inconsistent, and easily manipulated. FASTag, while successful for payment automation, does not address residency verification at all. This paper proposes SMART-TOLL (Spatial Memory and AI-assisted Residency Toll system) — an AI camera framework placed 2km before toll plazas that uses Automatic Number Plate Recognition (ANPR), behavioral frequency analysis, and vehicle color detection to compute a Residency Confidence Score (RCS) for every passing vehicle. The RCS algorithm fuses five weighted signals: monthly visit frequency, time-of-day consistency, number plate district geo-matching, directional travel patterns, and vehicle age/type from VAHAN. Vehicles scoring above 85 for three consecutive months are automatically whitelisted. The system also integrates VAHAN cross-checking for stolen/uninsured vehicles, a 5-color alert dashboard, emergency vehicle priority detection, a multi-toll district mesh network, and a Privacy-First ANPR (PF-ANPR) architecture compliant with India's DPDP Act 2023. Key contributions: the RCS algorithm, PF-ANPR architecture, and multi-toll behavioral mesh — none of which have been formally proposed for Indian toll systems in prior literature.