Real-time Traffic Analytics from Live Camara Feeds

点击翻译按钮获取中文摘要

The proposed system will detect and track vehicles, pedestrians and jaywalkers by using deep learning technique for real-time detection. Surveillance cameras will be installed at specific locations which will stream the real-time traffic data. A deep learning model will be built to detect vehicles and pedestrians and classify road users at high accuracy. The system not only allows users to create lanes and crosswalks to count vehicles and pedestrians but also creates jaywalking boundaries to detect jaywalkers in specific areas. The Deepsort algorithm will be utilized to identify and track the vehicles and pedestrians as well as their direction of movement. The system will allow users to add and manage multi camera streams (i.e., different angles) and features for specific cameras such as vehicle counting, pedestrian counting or jaywalking.The system illustrates various easy-to-understand dashboard visualizations to represent the multiple captured vehicles, pedestrians and jaywalkers in real-time. Furthermore, a warning system will be developed to alerts designated persons through SMS and/or email when an event of interest occurs. These insights can the system users to make informed decision for better traffic management safer roads.