<![CDATA[This study aims to develop algorithms of real-time traffic control system based on computer vision. In addition, the developed system is capable to analyze vehicles in a traffic stream at a traffic-controlled junction. The video recording sequence was installed near an intersection to control traffic lights during traffic congestion situations. This project's scope is limited to analyzing real-time traffic feeds and developing methods that count and track moving and stopping vehicles that approach a traffic junction. The project was designed using SIMULINK, which MathWorks created. Four algorithms were proposed to analyze the video signal inputs and estimate the number of vehicles detected. Gaussian mixture model and edge detection with frame differencing method were used to detect and track arrived vehicles. An optical flow-based approach was used to determine the number of stopped vehicles. Additionally, a vehicle classification algorithm was used to detect certain types of vehicles. In the Gaussian mixture model algorithm, implementing trained mask and geometric transform on each frame improved the perception of the outputs, which is defined by counting 1100 vehicles on the approach. Also, by using color detection, more control over traffic flow was obtained by prioritizing certain cars. The obtained results showed good representation of vehicle classification for the data detected in the developed system compared with the empirical data. The estimated errors were determined by achieving RMSPE < 15%, GHE < 5 and Um < 1.]]>
Copyrights © 2026
