In this paper, an algorithm is suggested in which vehicle detection is conducted using wavelet transformation and then traffic queue length in detection area is measured. Different from previous approaches such as edge detection method, background difference method, and spatial difference method, the algorithm is robust to noise and rapid changes in environment. For this reason, estimation of intensity of background image is not needed. In addition, the error rate caused by the headlight beam in nighttime can be drastically reduced. In the experiments with 80 samples in a standing queue in a single lane for daytime and nighttime, the error rate was 2.14% and 2.24%, respectively. For the moving queue the error rate for daytime and nighttime was 2.59% and 2.73%, respectively. (A*)
Abstract