On this page we are collected our research on detection of traffic flows through generic video
Vehicle Tracking System based on Videotaping Data
The relevance of microscopic traffic simulation models for the analysis of traffic operations and safety is based on their ability to reproduce real time traffic operations at a given location. Accordingly, one of the major steps in the development and application of these models is a good calibration of the model input parameters based on a thorough comparison of simulated and observed vehicle trajectories. Videotaping provides a low cost nonintrusive procedure for capturing individual vehicle operations over time, and as such, provides a useful tool for obtaining observational data for calibration and validation of traffic simulation models. Before videotaped data can be used, however, vehicle tracks or trajectories will need to be extracted from a frame by frame analysis of vehicle progression measurements concerning longitudinal and lateral position, speed, and acceleration over time. A system for tracking moving vehicles is presented that overcomes many of the practical limitations of current videotaping applications usually resulting from traffic and site conditions for the road segment being videotaped. The data extraction algorithm proposed in this paper provides a more flexible (less restrictive) method for videotaping of vehicle in the traffic stream, which attempts to overcome many of the limitations imposed by other more restricting taping methods currently in use. However, the accuracy of the vehicle tracking system needs to be assessed. This necessitates a comparison of individual vehicle trajectories as extracted from the video with the corresponding profiles obtained from baseline ground control points (GPS) referenced values for the same trajectories. This study has two objectives: 1) to introduce and describe the video extraction algorithm that allows vehicle segmentation and tracking, and the computation of traffic parameters from the tracking data, 2) to assess the accuracy of this algorithm with respect to changes in the horizontal viewing angle, and the number and placement of GCP along a given road segment. Preliminary results obtained from a case study shown that the number of GCP and the deflection angle from the perpendicular camera sightline to the roadway have a significant effect on the accuracy of the detected vehicle trajectories. Furthermore, the placement of GCP along the road segment has a significant effect on error, especially as it affects the scale of pixels at the edge of the video angle.
https://www.researchgate.net/publication/268217199_Vehicle_Tracking_System_based_on_Videotaping_Data
Evaluating the accuracy of a new algorithm for extracting vehicle tracking data from video taping
A methodology for tracking moving vehicles is presented that overcomes many of the practical limitations of current video taping applications many resulting from traffic and site conditions for the road segment being video-taped. The algorithm presented in this paper provides a sound inexpensive procedure for extracting vehicle tracking data with minimum video taping restrictions. This is achieved through a comprehensive filtering of videotaped images, removal of background distortions, reduced impact of image occlusion, identification and construction of blobs from pixel features, and an accurate link to fixed representative reference points inside of the video frame (Ground Control Points or GCP). The tracking algorithm has been applied to a sample of video-taped vehicle trajectories with parallel GPS geo-referenced information to investigated the effect of placement of GCP and video camera angle on error in vehicle tracking. The number of GCP and the deflection angle from the perpendicular camera sightline to the roadway have a significant effect on the accuracy of the detected vehicle trajectories. Slightly higher errors were noted for a small number of GCP. Accuracy in the tracking algorithm is important for the calibration and validation of microscopic traffic simulation models.
Safety performance measures: A comparison between microsimulation and observational data
Safety performance measures can be obtained either through simulation (based on well specified or calibrated traffic models) or experimentally through observational vehicle tracking data. Accurate calibration of traffic models ensures that simulated measures of safety performance are reflective of “real world” traffic conditions. The microscopic model, for a case study, allows the estimation of road safety performance through a series of indicators, representing interactions in real time between different pairs of vehicles belonging to the traffic stream. When these indicators reach a certain critical value, a possible accident scenario is identified. For the same case study, safety performance indicators are obtained through a video image processing algorithm for vehicle detection and tracking. The accuracy of the algorithm is evaluated with respect to GPS tracking measurements. The algorithm adopts a background subtraction-based approach for vehicle detection in 0.1 second increments. Since this approach is sensitive to background changes (or noise), a median filter technique has been introduced. Individual vehicles are detected and tracked using a region-based approach, whereby a connected zone (or blob) is assigned to each image, which is then tracked over time. In case of overlapping, where the designated blob may correspond to several vehicles, a real time sub-routine is accessed that manually discriminates each constituent vehicle’s specific position within the blob. Output from the algorithm application is expressed in terms of several trajectory descriptors over time, such as position and speed. The focus of this paper is on the analysis of road safety from two different perspectives: microsimulation and observational data. In this way it is possible to determine how microsimulation reflects “real” driver behavior and traffic conditions for a given case study.
An algorithm to acquire vehicle trajectories from video images
Several algorithms for the extraction of target trajectories through the video capture systems have been developed and implemented in commercial software systems, as Autoscope, Citilog and Traficon. These systems use a virtual loop approach can provide traffic information useful for the management and the control, but not for the detection of trajectories. Other systems, such as PEEK Video Trak-IQ and NGSIM-Video, using a tracking approach of vehicles and must be carefully calibrated before their application. All these systems have different problems during the detection of vehicle trajectories: measurement errors caused by congestion, by camera-shake, by partial or complete occlusion of the visual, by changes in lighting. The algorithms for tracking vehicle trajectories from video images can be classified into three categories: algorithms based on the recognition of areas (model based), algorithms based on the recognition of contours (contour based) and finally algorithms based on the recognition of the special features of ‘ image (feature based). Model-based algorithms provide accurate results for low traffic volumes or for specific types of vehicles. In the algorithms based on the recognition of areas, vehicles are recognized through uniform pixel blocks that can be monitored over time using cross-correlation measurements. This approach provides accurate results for non-congested traffic, but suffers from problems of identification of the vehicle when several vehicles can be assigned to the same block of pixels due to a partial overlapping of the areas (for example in case of congestion). The contour based algorithms are based on so-called models of the active contour (active contour), in which objects are detected through the analysis of the information the background image. Even this approach is affected by partial image coverage problems due to overlapping of the vehicles in the prospect of recovery. The algorithms based on the recognition of particular characteristics of the image (feature based) try to detect vehicles in specific points of the background and moving objects. This approach attempts to solve this problem in the two approaches described above. In this approach, in fact, even if it occurs the partial coverage of the vehicle, some of the characteristics of the vehicle remain recognizable and guarantee the uninterrupted monitoring.
Investigating safety issues in two-way rural highways
In the last years the growing need for a better mobility has coincided with an increase of congestion levels on transportation infrastructures and a consequent repercussion on safety aspects. For this reason researchers and technicians have focused on the study of safety performances on road networks identifying and applying all kinds of countermeasures useful to decrease accident risks. One of the most common methodologies to estimate safety makes use of inferential statistics applied to crash databases in what can be considered as a reactive approach to the problem. Although this method seems to intuitively link causes to effects, a good knowledge of the dynamics of the events preceding the crash may provide a more useful support to the implementation of appropriate countermeasures. Moreover, the problems of consistency and availability of crash data as well as the methodological challenges posed by the extremely random nature and the uniqueness of accidents have led to the development of complementary ap-proaches to improve road safety assessment, such as the observation of traffic conflicts and the use of microscopic traffic simulation. The potential of micro-scopic simulation in traffic safety has gained a growing interest due to recent de-velopment in human behavior modeling and real time vehicle data acquisition (Cunto and Saccomanno 2007, Cunto and Saccomanno 2008, Saccomanno et al. 2008, Yang et al. 2010, Cheol and Taejin 2010). A recent FHWA review of these models has concluded that: “when properly calibrated, traffic simulation models can provide useful and reliable information on individual driver responses to changing traffic and geometric road conditions” (FHWA, 2003). Since traffic con-ditions are inputs into safety performance, these models can also provide a solid causal platform for the study of safety.