Intelligent Traffic Management using Machine Learning for Peak and Off-Peak Hour Prediction
Traffic monitoring in cities is highly significant in transportation planning, in determining the extent of traffic conditions, and the operation of smart cities. However, conventional approaches, such as counting manually and relying on sensors installed in the infrastructure, have some significant issues. Manual techniques are also labor-intensive and difficult to maintain with time. Sensor-based systems, although automated, are expensive to install and maintain over time and these systems lack scalability. To avoid these issues, this study proposes an automated vehicle traffic analysis system, which applies deep learning and video analytics. Input video clips in the system are processed using the YOLO object detection model on a continuous live stream. Video frames are processed sequentially one after the other and combining the identified vehicles over time, in order to determine the density of the traffic and peak traffic periods, low traffic activity periods. This methodology involves live streaming recording, segmenting video into smaller clips, identifying frames, matching them in temporal sequence and statistical analysis to view traffic patterns. The data is also presented in the form of graphs and summary statistics to better demonstrate how the traffic flow varies with the time windows. The framework is easy to scale and affordable alternative to the conventional techniques, yet its functionality can be influenced by factors, such as video quality and the surroundings. Nevertheless, the experimental results show that the proposed methodology is competent when it comes to identifying the pattern of traffic activity based on the long-duration surveillance videos and, thus, makes it possible to make data-driven decisions in the smart transportation systems.