Computer vision and driver distraction: Developing a behaviour-flagging protocol for naturalistic driving data

Naturalistic driving studies (NDS) allow researchers to discreetly observe everyday, real-world driving to better understand the risk factors that contribute to hazardous situations. In particular, NDS designs provide high ecological validity in the study of driver distraction. With increasing dataset sizes, current best practice of manually reviewing videos to classify the occurrence of driving behaviours, including those that are indicative of distraction, is becoming increasingly impractical. Current statistical solutions underutilise available data and create further epistemic problems. Similarly, technical solutions such as eye-tracking often require dedicated hardware that is not readily accessible or feasible to use. A computer vision solution based on open-source software was developed and tested to improve the accuracy and speed of processing NDS video data for the purpose of quantifying the occurrence of driver distraction. Using classifier cascades, manually-reviewed video data from a previously published NDS was reanalysed and used as a benchmark of current best practice for performance comparison. Two software coding systems were developed – one based on hierarchical clustering (HC), and one based on gender differences (MF). Compared to manual video coding, HC achieved 86 percent concordance, 55 percent reduction in processing time, and classified an additional 69 percent of target behaviour not previously identified through manual review. MF achieved 67 percent concordance, a 75 percent reduction in processing time, and classified an additional 35 percent of target behaviour not identified through manual review. The findings highlight the improvements in processing speed and correctly classifying target behaviours achievable through the use of custom developed computer vision solutions. Suggestions for improved system performance and wider implementation are discussed.     

Driving with pets and motor vehicle collision involvement among older drivers: A prospective population-based study

ObjectiveDistracted driving is a major cause of motor vehicle collision (MVC) involvement. Pets have been identified as potential distraction to drivers, particularly in the front. This type of distraction could be worse for those with impairment in the cognitive aspects of visual processing. The purpose of this study is to evaluate the association between driving with pets and rates of motor vehicle collision involvement in a cohort of older drivers.MethodsA three-year prospective study was conducted in a population-based sample of 2000 licensed drivers aged 70 years and older. At the baseline visit, a trained interviewer asked participants about pet ownership, whether they drive with pets, how frequently, and where the pet sits in the vehicle. Motor vehicle collision (MVC) involvement during the three-year study period was obtained from the Alabama Department of Public Safety. At-fault status was determined by the police officer who arrived on the scene. Participants were followed until the earliest of death, driving cessation, or end of the study period. Poisson regression was used to calculate crude and adjusted rate ratios (RR) examining the association between pet ownership, presence of a pet in a vehicle, frequency of driving with a pet, and location of the pet inside with vehicle with any and at-fault MVC involvement. We examined whether the associations differed by higher order visual processing impairment status, as measured by Useful Field of View, Trails B, and Motor-free Visual Perception Test.ResultsRates of crash involvement were similar for older adults who have ever driven with a pet compared to those who never drove with their pet (RR = 1.15, 95% CI 0.76–1.75). Drivers who reported always or sometimes driving with their pet had higher MVC rates compared to pet owners who never drive with a pet, but this association was not statistically significant (RR = 1.39, 95% CI 0.86–2.24). In terms of location, those reporting having a pet frequently ride in the front of the vehicle had similar rates of MVC involvement compared to those who did not drive with a pet in the front. A similar pattern of results was observed for at-fault MVCs. This association was not modified by visual processing impairment status.ConclusionThe current study demonstrates a positive but non-significant association between frequently driving with pets and MVC involvement. More research is needed, particularly on restraint use and whether the pet was in the car at the time of the crash, to help characterize the public safety benefit of regulations on driving with pets.     

The influence of driver distraction on the severity of injuries sustained by teenage drivers and their passengers

Studies show that teenage drivers are at a higher risk for crashes. Opportunities to engage in technology and non-technology based distractions appear to be a particular concern among this age group. An ordered logit model was developed to predict the likelihood of a severe injury for these drivers and their passenger using a national crash database (the 2003, U.S. DOT-General Estimate System [GES]). As one would expect, speeding substantially increases the likelihood of severe injuries for teenage drivers and their passengers. The results of the analysis also reveal that teenage drivers have an increased likelihood of more severe injuries if distracted by a cell phone or by passengers than if the source of distraction was related to in-vehicle devices or if the driver was inattentive. Additionally, passengers of teenage drivers are more likely to sustain severe injuries when their driver is distracted by devices or passengers than with a non-distracted or inattentive driver. This supports the previous literature on teenage drivers and extends our understanding of injuries for this age group related to distraction-related crashes.     

Age-related differences in visual scanning at median-divided highway intersections in rural areas

The objective of this study was to examine age-related differences in visual scanning as drivers performed three separate maneuvers (going straight across, making a left and right turn) at two median-divided highway intersections with different crash frequencies. An on-road study was conducted with 60 drivers in three age groups: younger (18-25), middle-aged (35-55), and older (65-80). The study consisted of two between-subject (age and gender) and two within-subject variables (drive maneuver and intersection type). Drivers’ behavior was measured by the proportion of time they visually sampled towards the left, right and rearview mirror, and by an entropy rate representative of randomness in visual scanning. The results showed that older and younger drivers do not utilize their full scanning range when compared to middle-aged drivers, as indicated by lower entropy rate and the tendency to check fewer areas before executing a maneuver through the intersections. This trend was more obvious during left and right turn maneuvers indicating a greater likelihood to miss an unexpected event. Older drivers had a significantly smaller proportion of visual sampling to the left and right during intersection negotiations when compared to younger and middle-aged drivers. Older and younger drivers checked the rearview mirror significantly less when compared to middle-aged drivers.     

Examination of naturalistic driving practices in drivers with Parkinson's disease compared to age and gender-matched controls

A few studies suggest that drivers with Parkinson's disease (PD) may self-regulate or modify their driving behavior more than drivers without neurological disorders; however findings are limited to self-report. The purpose of this study was to objectively examine whether drivers with PD show more restrictive driving practices (exposure and patterns). Electronic devices were installed in the vehicles of 27 drivers with PD (71.6 ± 6.6; 78% men) and 20 matched controls (70.6 ± 7.9; 80% men) for two weeks and driving data were matched with aerial maps, weather and daylight archives and trip logs to examine driving context. Compared to controls, the PD group drove significantly less overall (number of trips, kilometres, duration), and proportionately less at night and on days with bad weather suggesting more restricted driving practices, congruent with lower ratings of driving comfort and abilities. However, they may not necessarily drive more cautiously or safely as they drove significantly faster (and over the speed limit) on highways and freeways and 19% reported driving problems over the two weeks. These preliminary findings need to be replicated and longitudinal studies using objective indicators are needed to examine changes in driving practices, as well as crash outcomes, as disease severity progresses.     

Driver distraction and driver inattention: definition, relationship and taxonomy

There is accumulating evidence that driver distraction and driver inattention are leading causes of vehicle crashes and incidents. However, as applied psychological constructs, they have been inconsistently defined and the relationship between them remains unclear. In this paper, driver distraction and driver inattention are defined and a taxonomy is presented in which driver distraction is distinguished from other forms of driver inattention. The taxonomy and the definitions provided are intended (a) to provide a common framework for coding different forms of driver inattention as contributing factors in crashes and incidents, so that comparable estimates of their role as contributing factors can be made across different studies, and (b) to make it possible to more accurately interpret and compare, across studies, the research findings for a given form of driver inattention.     

The efficacy of an educational intervention in promoting self-regulation among high-risk older drivers

Visual processing impairment increases crash risk among older drivers. Many older drivers meet the legal requirements for licensing despite having vision impairments that elevate crash risk. In this study, 365 older drivers who were licensed, visually-impaired, and crash-involved in the prior year were randomly assigned to an intervention group or usual-eye-care control group to evaluate the efficacy of an educational intervention that promoted the performance of self-regulatory practices. The educational curriculum was designed to change self-perceptions about vision impairment and how it can impact driver safety and to promote the avoidance of challenging driving situations through self-regulation, leading to reductions in driving exposure. Analyses compared the intervention and control groups at pre-test and 6 months post-test with respect to self-reported perceptions about vision and driving practices. At post-test, drivers who had received the educational intervention were more likely to acknowledge that the quality of their eyesight was less than excellent, report a higher frequency of avoiding challenging driving situations (e.g. left-turns) and report performing more self-regulatory practices (e.g. three right-turns) as compared to controls. Additionally, drivers in the educational intervention group reported significantly fewer days, fewer places and fewer trips made per week as compared to those not receiving the educational intervention. These findings imply that visually-impaired older drivers at higher risk for crash involvement may benefit from educational interventions by reducing their driving exposure and increasing their avoidance of visually challenging driving situations. A critical future step in this research program will be to examine whether this educational intervention has an impact on the safety of these high-risk older drivers by reducing their crash involvement in the years following the educational intervention.     

Reliability of drivers in urban intersections

The concept of human reliability has been widely used in industrial settings by human factors experts to optimise the person-task fit. Reliability is estimated by the probability that a task will successfully be completed by personnel in a given stage of system operation. Human Reliability Analysis (HRA) is a technique used to calculate human error probabilities as the ratio of errors committed to the number of opportunities for that error. To transfer this notion to the measurement of car driver reliability the following components are necessary: a taxonomy of driving tasks, a definition of correct behaviour in each of these tasks, a list of errors as deviations from the correct actions and an adequate observation method to register errors and opportunities for these errors. Use of the SAFE-task analysis procedure recently made it possible to derive driver errors directly from the normative analysis of behavioural requirements. Driver reliability estimates could be used to compare groups of tasks (e.g. different types of intersections with their respective regulations) as well as groups of drivers’ or individual drivers’ aptitudes. This approach was tested in a field study with 62 drivers of different age groups. The subjects drove an instrumented car and had to complete an urban test route, the main features of which were 18 intersections representing six different driving tasks. The subjects were accompanied by two trained observers who recorded driver errors using standardized observation sheets. Results indicate that error indices often vary between both the age group of drivers and the type of driving task. The highest error indices occurred in the non-signalised intersection tasks and the roundabout, which exactly equals the corresponding ratings of task complexity from the SAFE analysis. A comparison of age groups clearly shows the disadvantage of older drivers, whose error indices in nearly all tasks are significantly higher than those of the other groups. The vast majority of these errors could be explained by high task load in the intersections, as they represent difficult tasks. The discussion shows how reliability estimates can be used in a constructive way to propose changes in car design, intersection layout and regulation as well as driver training.     

Multiple driver distractions: A systemic transport problem

Strategies to contend with driver distraction may no longer be sufficient for the emerging variety of contemporary driver distractions. A more systematic and systemic approach holds promise for improved road safety but is not currently being developed. This systematic review of multiple driver distractions aims to address this gap and presents two key findings. Systematic classification of distracting tasks with respect to driving is challenging, and engagement with Multiple-Additional-to-Driving (MAD) tasks is almost universally detrimental to driving performance. A model is presented to assist in systematically characterising multiple driver demands. Identified literature is placed into context using the model and shortfalls are identified.     

Evaluating changes in driver behaviour: A risk profiling approach

New road safety strategies continue to be devised by researchers and policy makers with pay-as-you-drive (PAYD) schemes gaining increasing attention. However, empirically measuring the effectiveness of these strategies is challenging due to the influence of the road environment and other factors external to the driver. The analysis presented here applies Temporal and Spatial Identifiers to control for the road environment and Driver Behaviour Profiles to provide a common measure of driving behaviour based on the risk of a casualty crash for assessing the effectiveness of a PAYD scheme on reducing driving risks. The results show that in many cases personalised feedback alone is sufficient to induce significant changes, but the largest reductions in risk are observed when drivers are also awarded a financial incentive to change behaviour. Importantly, the more frequent the exposure to the speeding information, the greater the magnitude of the change. However, the changes are disproportionately associated with those that were already safer drivers in the baseline period suggesting that some drivers may be predisposed to changing their behaviour. These results suggest that it would be beneficial to provide real-time or daily feedback on speeding behaviour in conjunction with a financial reward scheme, potentially as a component of insurance premiums.     

Psychological factors for driver distraction and inattention in the Australian and New Zealand rail industry

A signal passed at danger (SPAD) event occurs when a train moves past a stop signal into a section of unauthorised track. SPAD events are frequently attributed to driver distraction and inattention, but few studies have explored the failure mode from the perspective of task demand and the ability of the driver to self-regulate in response to competing activities. This study aimed to provide a more informed understanding of distraction, inattention and SPAD-risk in the passenger rail task. The research approach combined focus groups with a generative task designed to stimulate situational insight. Twenty-eight train drivers participated from 8 different rail operators in Australia and New Zealand. Data were analysed thematically and revealed several moderating factors for driver distraction. Time-keeping pressure and certain aspects of the driver-controller dynamic were considered to distort performance, and distractions from station dwelling and novel events increased SPAD-risk. The results are conceptualised in a succinct model of distraction linking multiple factors with mechanisms that induced the attentional shift. The commonalities and inter-dynamics of the factors revealed insight into driving anxiety in the passenger rail mode, and suggested that SPAD-risk was intensified when three or more factors converged. The paper discusses these issues in the context of misappropriated attention, taxonomic implications, and directions for future research.     

The development of a naturalistic data collection system to perform critical incident analysis: an investigation of safety and fatigue issues in long-haul trucking

Traditionally, both epidemiological and empirical methods have been used to assess driving safety. This paper describes an alternative, hybrid, naturalistic approach to data collection that shares advantages with each traditional approach. Though this naturalistic approach draws on elements of several safety techniques that have been developed in the past, including the Hazard Analysis Technique, instrumented vehicle studies, and fleet studies of driving safety interventions, it has a number of unique elements. Sophisticated instrumented vehicles collected over 400,000 km of commercial vehicle data to address the long-haul trucking application described in this paper. The development of this data collection and analysis method and data collection instrumentation has resulted in a set of valuable tools to advance the current state-of-the-practice in driving safety assessment. An application of this unique approach to a study of long-haul truck driver performance, behavior, and fatigue is described herein.     

A before-and-after study of driver stopping propensity at red light camera intersections

This paper reports a before-and-after study which evaluated the impacts of installing and operating red light cameras at two "T" and one "X" signalized intersections on driver stopping propensity upon the onset of amber. Rather than using video cameras, a special purpose data logger working in conjunction with loop sensors was used to gather traffic parameters, vehicle stopping/crossing movements, and signal phases. Logistic modeling was employed to model the revealed stopping/crossing decisions of non-platoon vehicle drivers in response to the onset of amber with traffic, situational and behavioral variables, including their interactions. The results indicate that the variable Cam_Inst that gathered the impacts of red light camera (RLC) installation on driver decision-making at signalized intersection was significant at 5% level only for the camera approach model of the cross-intersection. The significance of Cam_Inst was undermined for the camera approaches at the T-intersections by the inclusion of a significant two-order variable defining the interaction of Cam_Inst with distance from the stop-line. One may, thus, infer that RLC has a fixed-quantum effect at the cross-intersection, but an accentuated effect with distance at the two T-intersections. Lastly, the effects of a RLC at an intersection on the stopping decision at the non-camera approaches were minimal.     

Health literacy of older drivers and the importance of health experience for self-regulation of driving behaviour

This study provides much needed information on the education level of older drivers regarding the impact of health conditions and medications on personal driving safety, where they source this information, and how this knowledge influences self-regulation of driving. Random and convenience sampling secured 322 Australian drivers (63.9% males) aged 65 years and over (M = 77.35 years, SD = 7.35) who completed a telephone interview. The majority of respondents (86%) had good knowledge about health conditions (health knowledge) and driving safety, however more than 50% was classified as having poor knowledge on the effects of certain medications (medication knowledge) and driving safety. Poorer health knowledge was associated with a reduced likelihood of driving over 100 km in adjusted models. Being older and having more than one medical condition was found to increase the likelihood of self-regulation of driving. Results indicate that health knowledge was less important for predicting driving behaviour than health experience. Of great interest was that up to 85.7% of respondents reported not receiving advice about the potential impact of their medical condition and driving from their doctor. The findings indicate a need for improved dissemination of evidence-based health information and education for older drivers and their doctors.     

Using naturalistic driving data to explore the association between traffic safety-related events and crash risk at driver level

There has been considerable research conducted over the last 40 years using traffic safety-related events to support road safety analyses. Dating back to traffic conflict studies from the 1960s these observational studies of driver behavior have been criticized due to: poor quality data; lack of available and useful exposure measures linked to the observations; the incomparability of self-reported safety-related events; and, the difficulty in assessing culpability for safety-related events. This study seeks to explore the relationships between driver characteristics and traffic safety-related events, and between traffic safety-related events and crash involvement while mitigating some of those limitations. The Virginia Tech Transportation Institute 100-Car Naturalistic Driving Study dataset, in which the participants’ vehicles were instrumented with various cameras and sensors during the study period, was used for this study. The study data set includes 90 drivers observed for 12–13 months driving. This study focuses on single vehicle run-off-road safety-related events only, including 14 crashes and 182 safety-related events (30 near crashes, and 152 crash-relevant incidents). Among the findings are: (1) drivers under age 25 are significantly more likely to be involved in safety-related events and crashes; and (2) significantly positive correlations exist between crashes, near crashes, and crash-relevant incidents. Although there is still much to learn about the factors affecting the positive correlation between safety-related events and crashes, a Bayesian multivariate Poisson log-normal model is shown to be useful to quantify the associations between safety-related events and crash risk while controlling for driver characteristics.     

Older drivers and rapid deceleration events: Salisbury Eye Evaluation Driving Study

Drivers who rapidly change speed while driving may be more at risk for a crash. We sought to determine the relationship of demographic, vision, and cognitive variables with episodes of rapid decelerations during five days of normal driving in a cohort of older drivers. In the Salisbury Eye Evaluation Driving Study, 1425 older drivers aged 67–87 were recruited from the Maryland Motor Vehicle Administration's rolls for licensees in Salisbury, Maryland. Participants had several measures of vision tested: visual acuity, contrast sensitivity, visual fields, and the attentional visual field. Participants were also tested for various domains of cognitive function including executive function, attention, psychomotor speed, and visual search. A custom created driving monitoring system (DMS) was used to capture rapid deceleration events (RDEs), defined as at least 350 milli-g deceleration, during a five day period of monitoring. The rate of RDE per mile driven was modeled using a negative binomial regression model with an offset of the logarithm of the number of miles driven. We found that 30% of older drivers had one or more RDE during a five day period, and of those, about 1/3 had four or more. The rate of RDE per mile driven was highest for those drivers driving <59 miles during the 5-day period of monitoring. However, older drivers with RDE's were more likely to have better scores in cognitive tests of psychomotor speed and visual search, and have faster brake reaction time. Further, greater average speed and maximum speed per driving segment was protective against RDE events. In conclusion, contrary to our hypothesis, older drivers who perform rapid decelerations tend to be more “fit”, with better measures of vision and cognition compared to those who do not have events of rapid deceleration.     

A preliminary study of sexual activity as a distraction for young drivers

In what may be the first in-depth study of sexual activity as a driving distraction in the US, a sample of 195 male and 511 female college students at a Midwestern university (M age = 19.7) participated in an on-line study of sex while driving (SWD). Of these, 64 (32.8%) men and 47 (9.3%) women had engaged in sex while driving (SWD). Nine percent of men and 29% of women had engaged in SWD as a passenger. In most recent SWD incidents, respondents reported that the two most common acts were oral sex (70.3%) and genital touching (60.4%). About 11% engaged in vaginal intercourse. Sexual activity lasted from 1 to 10 min for 42.7% of the respondents. Nearly half (49.1%) were traveling 61–80 mph during sex. Considering respondents’ lifetime incidents of SWD, the most common driving errors reported were speeding (37.8%), drifting into another lane (36%), and letting go of the steering wheel (10.8%). Only 1.8% nearly had a crash, and none actually had a crash. Separate regression analyses for male and female respondents revealed that lower intentions to engage in SWD in the future were associated with higher estimates of the probability of a car crash. The authors consider SWD to be an under-reported in-vehicle distraction and encourage more research and prevention efforts.     

Relationships between interest in motor racing and driver attitudes and behaviour amongst mature drivers: an Australian case study

Interest in motor racing is investigated as a possible source of influence on driver attitudes toward speeding and driver behaviour. Previous studies have identified links between motor racing and road accidents on public roads. One study found that the level of interest in motor racing was positively correlated with risky driving behaviours of young male drivers. This paper outlines a conceptual framework for investigating the association between interest in motor racing and speeding violations on public roads. A sample survey of households in Queanbeyan, NSW, was used to examine the relationship between the level of interest in motor racing and attitudes to speeding and driving violations in a group of mature drivers. Results indicate that the level of interest in motor racing is significantly related to attitudes towards speeding, controlling for age, education level and sensation seeking propensity. Higher levels of interest in motor racing are associated with higher pro-speeding attitudes. Unlike the previous research on young male drivers, there was no significant relationship between interest in motor racing and speeding violations for this study of mature drivers. The implications of the study for road safety interventions are discussed.     

Effects of practice, age, and task demands, on interference from a phone task while driving

Experimental research on the effects of cellular phone conversations on driving indicates that the phone task interferes with many driving-related functions, especially with older drivers. Unfortunately in past research (1) the dual task conditions were not repeated in order to test for learning, (2) the 'phone tasks' were not representative of real conversations, and (3) most often both the driving and the phone tasks were experimenter-paced. In real driving drivers learn to time-share various tasks, they can pace their driving to accommodate the demands of a phone conversation, and they can even partially pace the phone conversation to accommodate the driving demands. The present study was designed to better simulate real driving conditions by providing a simulated driving environment with repeated experiences of driving while carrying two different hands-free 'phone' tasks with different proximities to real conversations. In the course of five sessions of driving and using the phone, there was a learning effect on most of the driving measures. In addition, the interference from the phone task on many of the driving tasks diminished over time as expected. Finally, the interference effects were greater when the phone task was the often-used artificial math operations task than when it was an emotionally involving conversation, when the driving demands were greater, and when the drivers were older. Thus, the deleterious effects of conversing on the phone are very real initially, but may not be as severe with continued practice at the dual task, especially for drivers who are not old.     

Driver's lane keeping ability with eyes off road: Insights from a naturalistic study

Many studies have shown that driver inattention can influence lane-keeping ability. The majority of studies on lane keeping have been conducted in controlled on-road networks or in simulated environments. However, few studies have examined lane-keeping ability in naturalistic settings for the same purpose. In this current study, the relationship between driver inattention and lane keeping ability was examined using naturalistic data for 24 drivers. Driver inattention was placed into two categories based on whether drivers were looking forward toward the roadway (inattention with eyes-on-road) or not looking forward (inattention with eyes-off-road) while engaged in a secondary task. Repeated measures regression models were used to account for within-subject correlations. The results showed that, after accounting for driving speed and lane width, the eyes-off-road significantly increased the standard deviation of lane position (SDLP). The findings from this study are consistent with other studies that show that the amount of time drivers spend looking away from the road can impact drivers’ ability to maintain their lane position. Additionally, this paper demonstrates how driver inattention can be examined with real world data while accounting for the roadway, environment, and driver behavior.