Real-Time Pothole Detection System: A Deep Learning Approach with SSD
摘要
Roadway maintenance and safety are paramount concerns for commuters and transportation authorities, with potholes posing considerable risks to vehicles and causing accidents, damage, and increased maintenance expenses. Conventional pothole detection methods have been largely inefficient and lacking in real-time analytics, impeding proactive problem-solving. This study introduces an innovative, centralized platform for pothole detection and management, featuring an intuitive interface enriched with interactive tools. Utilizing a Single Shot Detector (SSD) ensures efficient and accurate pothole detection. Complementing the SSD, the Neo6M GPS Module and ESP32 Camera Module provide precise spatial mapping and data collection. Beyond detection, the platform integrates real-time mapping and route optimization, suggesting alternative pathways when damages are identified in-route. Additionally, the platform triggers automated email alerts to the concerned authorities every 24 h until the pothole is repaired, ensuring timely intervention. Furthermore, robust analytical tools allow users to sift through pothole data by criteria such as location and size. The results demonstrate effective pothole identification with varying detection accuracy across different configurations, highlighting the importance of optimal parameter tuning. The system enhances road safety by providing timely information and route suggestions, aiding in proactive road maintenance and improved commuter safety.