Within the context of the project Integral Approach of the Analysis of Traffic Accidents IAAV (in Dutch: Integrale Aanpak Analyse Verkeersongevallen), the main objective is the development of an integral multidisciplinary approach on the investigation of traffic accidents. TNO Human Factors conducted long-term video observations to collect data on the pre-crash phase of real accidents (what exactly happened just before the collision?). The video recordings of collisions were used to evaluate and validate the safety value of in-depth accident analyses, road scene analyses, and behavioural observations (including traffic conflicts). Long-term digital video recordings have been made at four urban intersections for a period between 19 and 22 months (24 hours/day), at an unsignalised yield T-junction, at two large-scale signalised intersections, and at a large-scale signalised T-junction. Unfortunately, due to the lack of a suitable automatic selection procedure of collisions, a time-consuming manual selection procedure by human observers had to be applied. In addition, events with deviant road user behaviour or potential conflicts were collected from video on a more or less random basis. Finally, for each intersection, one day has been selected to scan the video disks for potential conflicts. Collisions and conflicts have been analysed quantitatively using the VIDARTS (VIDeo Analysis of Road Traffic Scenes) approach and conflicts scored according to the criteria of the DOCTOR (Dutch Objective Conflict Technique for Operation and Research) technique. Quantitative measures included speed, acceleration, and interaction measures such as Time-To-Collision (TTC) and Post-Encroachment Time (PET). Results from the collisions and conflicts analyses have been confronted with the results of the separately conducted road scene analyses to evaluate the various approaches and to formulate recommendations for safety improvements at the four intersections under study. In total sixteen collisions could be identified from video, much less than was anticipated in the original project plan. Ten collisions were between two passenger cars, two single-vehicle accidents, two between a car and a bicyclist/moped rider, and two single road user accidents (one bicyclist and one scootmobiel rider). For each collision, a detailed description is given of the process just before and during the collision phase. In most cases, a plausible reason why people behaved as they did, could be deducted from the video images. The conflicts that were scored (either randomly during the collision scanning process or systematically for one day) clearly illustrated typical safety problems, and together with the peculiar events as collected by the human observers during the process of the selection of collisions, resulted in several observations about the typical lay-out and functioning of the intersection at hand. In general, it can be concluded that traffic conflicts and deviant behaviour, together with road scene analyses give a good insight in potential safety problems at specific locations such as intersections from a road users’ perspective, well in line with the results from the analysis of the collisions. With respect to the latter, remarkably, in most cases, another road user was (in)directly involved, either as a distracting or as a contributing element, for example by occluding the view of one of the road users involved. The sixteen video recorded collisions that were collected helped a lot to get a better insight in accident causation processes and to value the results of conflict observations and road scene analyses. This study once more made it clear that conflict observation and scoring the severity according to the criteria of the DOCTOR technique from video are feasible. It even has some advantages over observing conflicts directly at street level by trained observers as you can have repeated looks on what actually happened. Time-related measures such as Time-To-Collision (TTC) and Post-Encroachment Time (PET) are promising surrogate safety measures for analysing encounters between road users at intersections and certainly have potential to serve as predictors of accident risks, for example for the development and application of traffic safety modules in microscopic traffic simulation models. In this study, the video recordings of collisions were intended to be used as a means to validate the separate approaches of in-depth accident analyses, road scene analyses and behavioural observations (including conflicts), and not as an integral part of the IAAV method. However, the results that were obtained from it together with the further technological development to be expected for automatic video analysis techniques of real world traffic scenes, let us wonder whether this approach can be applied for collecting more naturalistic driving behaviour data and getting a better understanding of the processes of interactions among road users. Recently, this so-called site-based risk approach looking at all of the traffic passing through a given road segment is recognised as complementary to in-vehicle studies in the large scale programme SHRP-II (Strategic Highway Research Program II) in the United States. (Author/publisher) This publication may be accessed by Internet users at: http://www.ictct.org/workshop.php?workshop_nr=25
Abstract