Pothoscape: Survey of Automated Pothole Detection and Hazard Warning

Abstract

Potholes and other road-surface defects are persistent transportation-safety and infrastructure-maintenance problems. Conventional inspection methods can be labour-intensive, periodic and expensive, while isolated sensing approaches may detect a road anomaly without providing sufficient visual context, accurate location or timely warning to road users. This literature survey reviews research on automated pothole and road-damage detection using computer vision, deep learning, smartphone sensors, GPS, crowdsensing and integrated real-time systems. The reviewed studies show a progression from vibration-based and sensor-based detection toward image-based object detection and, more recently, systems that combine detection with geographic mapping and driver-oriented warning. The literature also indicates that model accuracy depends on dataset diversity, environmental conditions, camera quality and deployment constraints. Smartphone-based approaches offer low-cost sensing, whereas deep-learning approaches provide stronger visual identification but can require greater computational resources and representative training data. Based on the comparison, a practical gap remains in integrating low-cost automatic detection, location tagging, continuous road-condition mapping and timely hazard warning within one deployable workflow. The survey therefore establishes the research basis for Pothoscape, a proposed road intelligence system intended to detect potholes automatically, warn approaching users and maintain a continuously updated road-condition database for future maintenance support.