Entity

Time filter

Source Type


Huang G.,Shandong University | Xing J.,Shandong University | Meng L.,Shandong University | Li F.,Public Traffic General Company | Ma L.,Public Traffic General Company
ICSPS 2010 - Proceedings of the 2010 2nd International Conference on Signal Processing Systems | Year: 2010

The Bus Rapid Transmit (BRT) system is one of the rapid development forms of transit priority in China. In this paper, a probabilistic method is proposed to predict the travel time for BRT by analyzing historical traffic-data. The line 1 of BRT in Jinan, China, which has been equipped with GPS units, is selected to provide the collected data. The time of day is split into consecutive time intervals which will be supposed as subsets. For each of the subsets, expectations, variances and probability distributions of travel times exist, and demonstrate the dispersions and uncertainties. This method is put forward to present the probability distributions of travel time. The confidence intervals corresponding to the given confidence levels can be further obtained. Finally, the confidence intervals combined with the expectations of travel time, will be the feasible and informative predictions. © 2010 IEEE. Source


Ma Z.,Shandong University | Xing J.,Shandong University | Gao L.,Shandong University | Sha J.,Shandong University | Wu Y.,Public Traffic General Company
Advances in Intelligent and Soft Computing | Year: 2011

In this paper, an dynamic urban public transport passenger flow forecasting approach is proposed based on interact multiple model (IMM) method. The dynamic approach (DA) maximizes useful information content by assembling knowledge from correlate time sequences, and making full use of historical and real-time passenger flow data. The dynamic approach is accomplished as follows: By analyzing the source data, three correlate times sequences are constructed. The auto-regression (AR), autoregressive integrated moving average (ARIMA), seasonal ARIMA (SARIMA) models are selected to give predictions of the three correlate time sequence. The output of the dynamic IMM serves as the final prediction using the results from the three models. To assess the performance of different approaches, moving average, exponential smoothing, artificial neural network, ARIMA and the proposed dynamic approach are applied to the real passenger flow prediction. The results suggest that the DA can obtain a more accurate prediction than the other approaches. © 2011 Springer-Verlag Berlin Heidelberg. Source


Wang S.,Shandong University | Xing J.,Shandong University | Wu Y.,Public Traffic General Company | Xu W.,Shandong University | And 2 more authors.
Advances in Intelligent and Soft Computing | Year: 2012

A Double-Sources shortest path algorithm for urban road networks is proposed in this paper. In typical urban road networks, the probability that the ratio of the shortest path length to the Euclidean distance denoted by |SD| between source station and destination station is smaller than 1.414, is larger than 95%. Based on Dijkstra algorithm and the characteristics of the typical urban road networks, this algorithm starts at searching for the shortest path from the source station and destination station respectively and simultaneously and ends at having found all stations which are less than 0.702|SD| far from the source station or destination station. Compared to the single-source Dijkstra algorithm, theory analysis and experimental results both show that the algorithm can great reduce the time-complexity, especially on condition that stations in urban road networks uniformly distribute. © 2012 Springer-Verlag GmbH. Source


Zhenliang M.,Shandong University | Jianping X.,Shandong University | Shihao Y.,Shandong University | Yong W.,Public Traffic General Company | Yubing W.,Public Traffic General Company
Advances in Information Sciences and Service Sciences | Year: 2011

Accurate bus arrival time prediction is crucial to the development of intelligent transportation systems and advanced traveler information systems. Automatic stop announcement (ASA) systems have been implemented in various public transit systems to realize automatic stop broadcasting along with other information such as location, travel time, etc. Such information has great potential as input data for a variety of applications including operations management, service planning, and performance evaluation. In this study, an aggregation method is developed for dynamic bus arrival time prediction, using data collected by a real world ASA system. The method implements in two stages, firstly bus baseline estimate travel time along each road section is calculated using SVM with historical trip data at given time-of-day, day-of-week and weather conditions. The second one is a H∞ filter based dynamic algorithm to adjust the arrival time prediction using the most recent trip information and SVM output. The method enables the prediction more accuracy and robust by taking into account of both historical data and real-time information, and there is no assumption of the noise, which is assumed to be Gaussian white by Kalman method. Experiments show that the aggregation method is quite powerful in bus arrival time prediction by comparing with the corresponding SVM model and SVM-KF algorithm. Source


Meng X.,Shandong University | Meng X.,Public Traffic General Company | Xing J.,Shandong University | Xing J.,Public Traffic General Company | And 8 more authors.
Advanced Materials Research | Year: 2012

To detect moving pedestrians in video surveillance rapidly and accurately, the pedestrian detection process is divided into two stages: motion detection stage and classification stage. In the motion detection stage motion detection is exploited to rapidly get the motion regions. By getting these interesting regions which may contain pedestrians, the regions used for next stage are greatly reduced and the whole pedestrian detection speed is improved. In the classification stage HOG features are used to classify interesting regions into pedestrian region and non-pedestrian region. In the motion detection stage, an improved frame differential method is used. The threshold changes with the pixel value changes and can adapt to the environment change, such as background difference, lightness change and motion speed. In the classification stage, an improved HOG method is used. Multi-scale HOG features are used to get better performance and Adaboost algorithm is used to train cascaded classifier for a higher detection speed. To evaluate the performance of our method, we test it on several video sequences taken from different scenes. Experiment results show that our method achieves good performance with an average over 90% detection rate. And our method can process over 20 frames per second and satisfies the demand for real-time task. © (2012) Trans Tech Publications, Switzerland. Source

Discover hidden collaborations