Y.L. Ma, R. Zhang, Z.F. Zhang, J. Gao

Pages: 55-62

Abstract
Traffic flow calculation has a great significance on transport planning, traffic congestion relief and traffic accidents prevention. This paper proposes a video-based extended Gaussian mixture model (EGMM) to calculate traffic flow of road segment. This method first use EGMM to detect the moving vehicles. Then the detected vehicles are calculated for traffic flow. The new model fuses GMM and frame difference, which can enhance the demarcation point between background and foreground. Experiments in different weather (sunny and rainy) and different time (day and night) have been made. They can prove that the new model has been great robustness in weathers and road conditions. Traffic flow can also be calculated correctly and effectively in the situation with various complicated factors.

Keywords: traffic flow; extended Gaussian mixture model; frame difference