Le Xin

A robust vehicle queuing and dissipation detection method based on two cameras

This paper proposed a robust framework of detecting the real-time queuing and dissipation of a vehicle queue by two cameras, one fixed at the front of the stop line and the other somewhere behind the stop line, jointly monitoring the interested region with opposite and long-range views. Firstly, the position changes of the tail and head of a vehicle queue, which accurately describe the formation and dissipation of the queue, can be efficiently tracked in each camera at intersection during morning and evening rush hours, with a duplex flexible window fused with the Haar feature based AdaBoost cascade classifiers. Secondly, the data of these two cameras in this large-area outdoor traffic application are fused at decision level to improve the accuracy of the tracking, according to the tracking result in each camera. Then, the queue length and stop delay of vehicles can be calculated readily. Experiments show that the proposed method can detect the formation and dissipation of the queue under varying illumination in real time, and that the accuracy rate is about 90.24%. Therefore, this method can be further applied to traffic congestion monitoring and traffic signal controlling.

Lane Detection in Video-Based Intelligent Transportation Monitoring via Fast Extracting and Clustering of Vehicle Motion Trajectories

Lane detection is a crucial process in video-based transportation monitoring system. This paper proposes a novel method to detect the lane center via rapid extraction and high accuracy clustering of vehicle motion trajectories. First, we use the activity map to realize automatically the extraction of road region, the calibration of dynamic camera, and the setting of three virtual detecting lines. Secondly, the three virtual detecting lines and a local background model with traffic flow feedback are used to extract and group vehicle feature points in unit of vehicle. Then, the feature point groups are described accurately by edge weighted dynamic graph and modified by a motion-similarity Kalman filter during the sparse feature point tracking. After obtaining the vehicle trajectories, a rough -means incremental clustering with Hausdorff distance is designed to realize the rapid online extraction of lane center with high accuracy. The use of rough set reduces effectively the accuracy decrease, which results from the trajectories that run irregularly. Experimental results prove that the proposed method can detect lane center position efficiently, the affected time of subsequent tasks can be reduced obviously, and the safety of traffic surveillance systems can be enhanced significantly.

Traffic flow characteristic analysis at intersections from multi-layer spectral clustering of motion patterns using raw vehicle trajectory

Road intersections are important components of urban road system. It is the traffic flow characteristics representing the current traffic situation that provide a basis for the planning, designing and management of intersections. In this paper, we constructed an automatic processing framework on traffic flow characteristics analysis and understanding the traffic state at the urban road intersections, based on the collected raw vehicle motion trajectories. Our proposed method is basically attributed to identifying distinct vehicle motion patterns at intersections hierarchically using raw trajectory. Firstly, the fundamental assumption in traditional approaches that the trajectory set of high quality are readily available after manual rectification is not taken for granted any more. And by fully analyzing the local characteristics of trajectories, we figure out and explain various patterns behind traffic flow as well as yielded higher accuracy in motion trajectory clustering under the multi-layer spectral clustering method. At last, coupling the analyzing results with the surrounding characteristics of the intersection, the examples computing traffic flow features and predicting vehicle activity illustrates the potential of applying vehicle trajectories to traffic study, which are all suggested by experimental results.

High-efficient detection of traffic parameters by using two foreground temporal-spatial images

Real-time detection of vehicular volume, mean speed and vehicle type has important significance, but the existing video-based detection methods are not satisfactory at processing speed and accuracy. This paper proposes a high-efficient method to detect all the three parameters from two foreground temporal-spatial images (TSIs) directly, which are obtained from two virtual detection lines (VDLs) in video frames. Such usage of the TSIs provides a feasible approach to solve the problems of vehicle occlusion, mean-speed estimation, and vehicle classification without using original frame images. Firstly, for improving the accuracy of detection, during generation of the foreground TSIs, we set a small-wide region of interest for each VDL and propose a local background subtraction method and an improved moving shadows elimination method to eliminate unwanted interferences. Then, in order to reduce the calculation complexity, during extraction of the parameters, we analyze the feasibility of vehicle classification direct from the foreground TSIs, and propose a method to extract shape-feature vector from the TSIs directly. The dependence on original frame images is minimized, so the pressing speed is improved obviously. Experimental results prove the feasibility and efficiency of the proposed method.

A robust detection method of vehicle queue and dissipation during evening rush hour

By vehicle presence detection and movement analysis on the monitored region based on video processing technology, a real-time robust detection method of vehicle queue and dissipation is proposed in this paper. With a duplex flexible window fused with the Haar feature based AdaBoost cascade classifiers, the method can efficiently track the position changes of the tail and head of a vehicle queue at intersection during evening rush hour, which accurately describes the formation and dissipation of the queue. Then, the queue length and stop delay of vehicles can be calculated. Experiments showed that our method can real-time track the formation and dissipation of the queue during the evening rush hour when the illumination varies greatly from light to dark, and that the accuracy rate is about 91.54%. Therefore, this method can be further applied to traffic congestion monitoring and traffic signal controlling.

Real-Time Detection and Tracking of Traffic Shock Waves by Conjugated Low-Angle Cameras

This paper proposes a method to detect and to track in real time three traffic shock waves (i.e., queuing, discharge, and departure) that appear in a section at a signalized intersection. Detection and tracking of the shock waves are accomplished by conjugated low-angle cameras, one of which is installed in front of the stop line and the other at a proper place behind the stop line. The cameras jointly monitor the regions of interest in the section with opposite and long-range views. On the basis of the weighted least squares method, the video data captured by the conjugated cameras are fused at the pixel level. Then the fused images are adapted to track the queuing and discharge shock waves in real time by a duplex, flexible window algorithm. Simultaneously, the Haar feature–based AdaBoost cascade classifiers are adopted to identify the vehicle tails and heads and to adjust the tracking results in the flexible window. To track the departure shock wave in real time, the discharge speed of vehicles is obtained through the combined tracking of sparse feature points of the vehicles in the regions of interest of the conjugated cameras. Experimental results show that the proposed method can track the traffic shock waves accurately in real time under changed light conditions during evening rush hour. The tracking results of the shock waves can be further applied to obtain various traffic parameters (e.g., queue length, stop delay, discharge time).

Detecting and locating of traffic incidents in a road segment based on lane-changing characteristics

ABSTRACT In road traffic, stalled vehicles, spilled loads and traffic accidents often result in traffic jams or secondary accidents. It is better for us to detect and locate these non-recurrent traffic incidents accurately as soon as possible, so that early warning, timely incident handling and speedy congestion evacuation will be better achieved. This paper proposes an automatic detecting and locating method of traffic incidents in a road segment based on lane-changing characteristics. At regular time intervals, for each traffic lane, the method calculates the lane-changing-out ratio of vehicles and judges whether a traffic incident has happened. For a traffic-incident lane, entry points and exit points of lane-changing vehicles are considered as positive and negative samples, respectively, then the traffic incident is confirmed further and its position is calculated by searching for the optimal separating surface of the two kinds of samples. Experimental results prove that the method is practicable and effective.

Calculating Vehicle-to-Vehicle Distance Based on License Plate Detection

Vehicle-to-vehicle distance calculation has a great significance to driving assistance and estimation of traffic condition. In this paper, we present an on-board video-based method about calculating distance gap. The method is mainly divided into three major stages. At first stage, an Adaboost cascade classifier using Haar-like features of sample pictures is used to detect preceding vehicles. At second stage, a fusion algorithm combining Maximally Stable Extremal Regions (MSER) for far vehicles with vertical texture method for close vehicles is applied to locate license plate. At the third stage, distance gap is calculated according to the pixel height of plate and the proportion of plate pixel height. Experimental results in this paper showed excellent performance of the method in calculating distance gap.

Extraction method of traffic parameters based on detecting traffic waves at intersections

Extraction of traffic parameters, such as queue length, stop delay and etc, of a road section between intersections is significant for traffic signal control. It is a challenge to get high precision parameters by video processing technologies during peak hours, for the serious vehicle queue. We propose a method of acquiring traffic parameters based on detecting traffic waves at intersections. We use information techniques to fuse the images from two cameras to detect and track traffic waves. On this basis, the various parameters are calculated with high precision. Experimental results show that the method has better performance during different peak periods.