In intelligent transportation systems, there are many tasks that rely on the detection of road congestion, such as traffic signal scheduling and traffic accident detection. As traditional methods for traffic congestion detection are difficult to use, expensive, and may cause damage to the road surface, this paper presents a method for road congestion detection that is based on multidimensional visual features and a convolutional neural network (CNN). This method first detects the density of foreground objects by using a gray-level co-occurrence matrix; second, the speed of moving objects is detected by using the Lucas-Kanade optical flow with pyramid implementation. Third, a Gaussian mixture model is used to model the background, and the CNN is then used to accurately detect the final foreground from the candidate foregrounds. Finally, the proposed method performs road congestion detection in terms of a multidimensional feature space, including traffic density, traffic velocity, road occupancy, and traffic flow. Furthermore, we propose an information entropy method using a histogram of optical flow to enhance the accuracy and reliability of road congestion detection. Simulation results via quantitative and qualitative assessment indicate that the proposed method is able to significantly outperform the state-of-the-art road-traffic congestion detection methods due to the fusion of multidimensional features using the CNN.

Considering both the tracking safety of multi-HSTs and the operational efficiency of a single HST, intelligent safe driving methods (ISDMs) are proposed to obtain better speed-distance curves by integrating hybrid automata (HA) with data mining algorithms in this paper. To begin with, an intelligent safe distance controller is established by using HA to ensure the tracking safety of multi-HSTs' operation in real time. Then, data-driven intelligent driving methods based on ensemble algorithms (Bagging or Adaboost.R) and classification and regression tree (CART) are proposed to discover the potential driving rules from the field driving data. Furthermore, because of the continuous rise of HST's operation mileage, the driving data collected from HST has increased tremendously compared with the subways. So, an iterative pruning error minimization algorithm is designed to reduce the redundancy of the driving data and improve the computational speed of the learning process. Finally, compared with the automatic train operation (ATO) method, the energy consumption of B-CART, A-CART, and S-A-CART algorithms can be decreased by 3.32%, 3.80%, and 4.30%, respectively.

For high-speed train (HST), high-precision of train positioning is important to guarantee train safety and operational efficiency. For improving train positioning accuracy, we develop a mathematical positioning model by analyzing the wireless position report created by HST. To begin with, k-means algorithm is integrated with the least square support vector machine (LSSVM) to differentiate the position data and establish the corresponding prediction model for each position data class. Then, the ant colony optimization (ACO) algorithm is introduced to adaptively optimize the clustering number of position data and solve the over-fitting problem of the single k-means algorithm. So, a better classification of position data can be obtained by ACO-k-means than the single k-means algorithm. Furthermore, the online learning algorithms are designed for improving the adaptability and real-time performance of established positioning model. Finally, the field data of Beijing-Shanghai high-speed railway (BS_HSR) is used to test the performance of the established positioning models. Experiments on real-world positioning data sets from BS_HSR illustrate that the proposed methods can enhance the real-time performance in online updating process on the premise of reducing the positioning error.