Automated traffic counting data collection and analysis

Low, Anand Hong Ren (2021) Automated traffic counting data collection and analysis. Final Year Project, UTAR.

Preview

PDF Download (2240Kb) | Preview

Abstract

The increase in the number of vehicles purchased over the years cause a high volume of vehicles on the road. This leads to traffic congestion especially in urban areas. This problem disrupts the daily life of many people. It is important to conduct traffic analysis and surveys to extract traffic information which would be useful for solving and evaluating the quality of transportation. Optimal traffic arrangements that reduce traffic congestion can be designed by engineers using the collected traffic data. Traffic data collection is also useful for other issues such as vehicle accidents, managing parking areas, speeding, vehicle theft detection and others. There have been many methods of traffic data collection proposed and implemented over the years, each with their own pros and cons. This project proposed an automated traffic counting data collection and analysis algorithm that is able to use computer vision to count and measure the speed of vehicles, while also able to classify vehicles into different categories using the power of deep learning and AI. The performance of the algorithm is determined by the counting, classification, and speed measuring accuracy. The factors affecting the performance of the algorithm is discussed. The system is able to performance the tasks when it is in the bright condition with the accuracy of more than 95%. However, the accuracy is dropped to 50% when the condition is dark. This is due to the system is unable to detect the vehicle in such condition.

Actions (login required)