主催: Eastern Asia Society for Transportation Studies
p. 328
An algorithm to track vehicles and determine traffic flow parameters through images generated by a surveillance camera in real time is developed. It aims to obtain traffic parameters such as vehicle count, vehicle speed, vehicle time headways, traffic volume and vehicle classification from a complex traffic scene. Two different video vehicle detection algorithms were developed for two different image types in processing, namely grey scale image technique and colour image technique. The vehicle is detected due to the difference of pixel value with the threshold value. Vehicle classification employed in this project is length based classification. Both methods give a high accuracy in vehicle detection and classification result. From the experiments which are done in real time, the accuracy of traffic parameters obtained is more than 80%. The develop algorithms may can be applied to various site locations with different luminance and lighting condition.