A dual-IMU based adaptive road anomaly detection and mapping system with privacy-preserving visual verification

Abstract

The present invention is related to a dual-IMU based adaptive road anomaly detection and mapping system with privacy-preserving visual verification. The invention discloses a low-cost, fleet-deployable road anomaly detection system utilizing a dual-IMU architecture to distinguish true road defects from vehicle-induced vibrations. A chassis-mounted IMU and a suspension-mounted IMU work in tandem with an adaptive calibration mechanism employing rolling median filtering and dynamic bias adjustment to minimize false positives across different vehicle types. An event-triggered rear-facing camera is selectively activated upon verified anomaly detection, thereby preserving privacy and reducing bandwidth. On-device TinyML classification enables lightweight anomaly recognition without dependence on smartphones or cloud-only processing. Verified anomaly data is securely transmitted via MQTT over TLS or HTTPS and aggregated into a GIS-based dashboard for visualization and decision-making. The invention enables scalable, privacy-preserving, and accurate road quality monitoring suitable for smart cities, fleet operators, and public authorities.