Compared with the traffic flows on freeways, the traffic in urban networks is significantly influenced by signal controls. Therefore, signal timing parameters are required in lots of urban research topics, including signal performance measurements and traffic control methods. However, direct access to signal timing information is prohibitively difficult due to the diversities in signal machines and signal schemes. The objective of this paper is to demonstrate the feasibility of estimating signal timing parameters using point detectors. We use the passing timestamps obtained from ANPR cameras as an example data source. Firstly, a robust method is proposed to determine whether a new cycle starts by recognizing cycle breaking vehicles. Then cycle times, starts of greens, and changes in signal schedules are deduced by analyzing the periodic features of cycle breaking vehicles. Finally, a sensitivity analysis is carried out to show the influence of penetration rates.

Travel time distribution is an essential tool for measuring urban traffic performance, a subject that has been studied for decades. This paper conducts a comprehensive investigation of two types of travel time distributions using extensive license plate recognition data from Automatic Number Plate Recognition techniques on four signalized arterials in Guiyang, China. The travel time plane distributions presented in the overlay charts of observed travel times usually exhibit significant stratified data strips. When considering signal schemes, we observe that the cycle times of the first upstream and last downstream intersections are the determining factors for the data patterns of travel time plane distributions. We also investigate the characteristics of single or multiple peaks within various departure time windows. The results indicate that travel time statistical distributions are more likely to exhibit multiple states under short time windows. As for the shapes of travel time statistical distributions, skewness and kurtosis are used as descriptive statistics. The results show that the majority of statistical distributions are positively skewed and leptokurtic, and the skewness is highly correlated with the kurtosis. A stable skewness and kurtosis at a relatively lower level may be caused by lower travel time reliabilities.