VARGAS-MARTÍN, FERNANDO PhD; GARCÍA-PÉREZ, MIGUEL A. PhD
Opaque objects block a visual field region that is larger the nearer the object is to the eyes. New users of prescription glasses complain of the view of the spectacle frame, which causes a visual field occlusion whose implications for driving have been quantitatively assessed.1 This occlusion occurs in the far periphery and its consequences are easy to circumvent with head movements.
Potentially more dangerous occlusions are caused by opaque objects near the eyes but not mounted on the head, objects such as window posts or other elements within the cockpit of aircraft or in the interior of vehicles. Because of their position, the occlusion that these elements cause cannot be circumvented with head movements, thus exerting a permanent blockage that may eventually threaten safety. For instance, Roscoe and Hull2 determined that oversized window posts in the cockpit of the DC-9 aircraft must have prevented the pilot of Inex Adria Airways flight 500 from seeing a British Airways Trident 3 aircraft that had been on a stationary collision course with his DC-9 for almost 3 minutes over Zagreb (in the former Yugoslavia) on a clear morning in 1976, resulting in a midair collision that killed all 176 passengers and crew aboard both aircraft. This was indeed the fourth midair collision in which a DC-9 crew should have been able to see the other aircraft at least momentarily before the crash, but the transmissions between the DC-9 crew and the Zagreb air traffic controller (whose radar set did not receive the code and altitude of a DC-9 whose transponder was on standby) clearly revealed that the DC-9 crew did not see the Trident 3 until less than a second before the crash.
A number of car accidents are attributed to what has been termed “looked-but-failed-to-see errors” because the driver had actually looked in the direction where the other parties were but failed to see them. Some of these accidents may indeed have been caused by visual field occlusions, but the only direct evidence that occlusions caused by elements of the vehicle are acknowledged by drivers as potential threats to safety comes from a survey3 in which “11/29 drivers said that the A-pillara sometimes or often restricts their vision out of the car and 14 stated that they sometimes or often had to move their head to see round it,” further noting that “two drivers reported near misses that they had experienced as a result of A-pillar obscuration. One had had several near misses at T-junctions in several cars and the other had failed to see an approaching car at a roundabout.” A number of studies have analyzed the cause of looked-but-failed-to-see errors,4–8 and all of them have proposed explanations based on drivers' inattention or inappropriate visual search strategies. None of these studies considered occlusions caused by opaque elements inside the automobile (e.g., A-pillars or the inside rearview mirror) as contributing factors. Yet, these elements occlude relevant parts of the visual field when a driver scans the left or the right leg of an intersection while judging the need to give way while in slow motion. Current classification systems for the cause of accidents9,10 only consider external visual obstacles and do not mention any element of the automobile that the driver is on as an obstacle to vision. Elucidating the role of these elements in car accidents will be useful for a number of reasons, including the proposal of recommendations for automobile design, for modifications of road environments, or for the placement of road signs and street furniture.
All things considered, obstacles to vision within the automobile that the driver is on may have a role in accidents caused by looked-but-failed-to-see errors, but the issue has never been investigated and, thus, unequivocal evidence has never been reported. Investigating this issue faces the problem that there is no easy way to determine the location of visual field occlusions when at the wheel. At the same time, and despite recommendations that this research be carried out,11 no detailed study appears to have been published on the actual visual field of a driver at the wheel.12,13 The goal of this article is to describe a technique that can be used for measuring the available visual field of a driver at the wheel. We also illustrate the potential of this technique by reporting preliminary results of its application, including static visual field recordings at the wheel while looking both straight ahead and also sideways as if to make a turn, and dynamic recordings during natural driving.
Subjects and Automobiles
Three subjects participated in the study, all of whom are experienced drivers regularly driving more than one car. Subjects 1 to 3 were, respectively, 42, 34, and 35 years old; their height was 181, 169, and 172 cm; and their interpupillary distance was 65, 72, and 60 mm. Three automobiles (see Fig. 1) were used in this study: a 2003 Citroën Xsara, a 2002 Citroën Xsara Picasso, and a 1988 Seat Ibiza.
Snapshots used to determine visual fields were taken with a 2005XA B/W minicamera (RF Concepts Ltd.; www.rfconcepts.co.uk), which has a 1/3" EXview HAD CCD SONY image sensor. Video recordings during driving were taken with a wireless GD-LD-208C color minicamera (HungTek Technology Co.; www.kinghung.ebigchina.com), which has a 1/3" CMOS image sensor. Both cameras were fitted with wide-angle V-4301.9 to 2.0 FT lenses (Marshall Electronics Inc.; www.mars-cam.com), and we checked that both cameras have identical fields when fitted with this lens. The cameras were clipped onto slim plastic frames (see Fig. 2). Subjects wore this apparatus along with their usual spectacle frames, and the camera provided external composite video input for an MV600i Canon digital video camera in which snapshots and video clips were recorded.
Camera View Calibration
We recorded visual fields with a method similar to that described by Vargas-Martín,14 which requires relating a camera view to a subject's visual field. The apparatus was mounted on a stand and placed such that the entrance pupil of the lens was 24.5 cm away from a translucent screen onto which the linear field in Figure 3A was rear-projected. The field displayed concentric circles spaced every 10° of visual angle for an observer whose eyes were on either side of the location of the camera and also included radial lines every 15° around the circle.
A snapshot (Fig. 3B) was thus taken, which represents the camera view of an observer's visual field. A cleaned-up version of this image (Fig. 3C) served as a measurement grid. Note that the outer circle has a radius of 70° of visual angle and, thus, it is only slightly larger than the typical binocular visual field of a human observer (see Fig. 8.17 in reference 15).
Visual Field Recordings at the Wheel
Each subject wore the spectacle-mounted camera and sat at the wheel adjusting the seat for comfortable driving, although the automobile was parked throughout these recordings. The automobile was facing a wall and at a distance such that the driver's eyes (after the seat had been adjusted by the driver) were 2.4 m away from the wall, as determined by pushing the automobile until the driver's eyes met a laterally projected vertical laser line. The driver then moved his or her head slightly backward until the laser line met the entrance pupil of the camera lens, so that scenes recorded from inside the automobile matched the geometry used to obtain the measurement grid in the laboratory. Three locations had been marked on the wall approximately at the driver's eye height. One of these locations was directly straight ahead and the two other locations were 3 m (51.3°) to the left and to the right of the central location. Because the automobile blocks different parts of the visual field when drivers turn their gaze to fixate on each of these locations (as is the case when making sharp right or left turns), binocular visual fields were recorded by taking a snapshot with the spectacle-mounted camera while the subject's head was oriented toward these locations.
The recorded scene was then digitally overlaid with the measurement grid, and the part of the grid that falls outside zones occluded by opaque elements in the interior of the automobile thus represents the available visual field of the driver at the wheel.
Subjects drove in the city of Murcia, Spain, on each of the automobiles while wearing the spectacle-mounted camera, taking a path involving sharp left and right turns at narrow intersections to gather actual data on the external information that the automobile allows into the driver's visual field. Videos were also recorded while driving on a nearby winding road.
Occlusions Caused by Different Automobiles
It should be borne in mind that a driver does not maintain a fixed head position during natural driving, and that our measurements with parked automobiles were taken with the drivers' preferred head position after they had adjusted the seat for comfortable driving. Consequently, the visual field occlusions that we report next do not represent permanent loss of visual information at the precise angular locations shown in our figures. Yet, once they are corrected for binocularity, the width of these occlusions certainly indicates the extent of the visual field that will be blocked at a given time. We will not apply any correction for binocularity in the present report.
Figure 4 shows the visual field of subject 1 when gazing to the left, straight ahead, and to the right while sitting at the wheel on each of the automobiles (rows). For a given direction of gaze, there are clear differences among automobiles, as discussed next.
When gazing to the left (left column in Fig. 4), the vertical extent of the visual field at the point of gaze is symmetric and spans approximately ±35° both in the Citroën Xsara (top picture) and in the Seat Ibiza (bottom picture), but it is markedly asymmetric and spans only from approximately 15° above to approximately 35° below the point of gaze in the Citroën Xsara Picasso (center picture). In addition, for this direction of gaze, the near edge of the left A-pillar falls 30° to the right of the line of gaze in the Seat Ibiza (bottom picture), 20° in the Citroën Xsara (top picture), and <10° in the Citroën Xsara Picasso (center picture). This A-pillar blocks a region of the visual field whose width along the horizontal meridian differs also across automobiles: only approximately 8° of visual angle in the Seat Ibiza (bottom picture), approximately 12° in the Citroën Xsara (top picture), and more than 20° in the Citroën Xsara Picasso as a result of its split A-pillar (center picture).
When gazing straight ahead (center column in Fig. 4), the visual field is severely curtailed in all automobiles, spanning only approximately ±15° vertically at the point of gaze. The left A-pillar blocks a region that is centered around 30° in the periphery, and has a span along the horizontal meridian of approximately 10° in the Citroën Xsara and the Seat Ibiza (top and bottom pictures) compared with more than 20° in the Citroën Xsara Picasso (center picture). In all automobiles, the inside rearview mirror blocks a region that is just above the horizontal meridian and covers a horizontal range of approximately 20°.
When gazing to the right (right column in Fig. 4), there is barely a clear circular area around fixation with a radius of 10°. The inside rearview mirror and the right A-pillar limit visibility to the left and to the right of fixation, and the roof and dashboard restrict the vertical span of the visual field. Note also that an occupant on the passenger seat will further block the right side of the visual field of the driver. For this direction of gaze, differences among automobiles are minor compared with the major curtailment caused by the mere presence of these opaque elements.
Visual fields for the two other drivers showed similar characteristics, although the precise location where the occlusions occur varied, as described next.
Occlusions Caused on Drivers of Different Height
To better compare the visual fields of different drivers, the photographs in Figure 4 as well as analogous photographs for the remaining drivers were used to obtain line sketches of the available visual field of each driver on each automobile for each direction of gaze. These sketches were drawn by hand on enlarged versions of the original digital photographs. Each panel in Figure 5 superimposes the resultant sketches for all drivers on each automobile (rows) and for each direction of gaze (columns).
By comparison with the common measurement grid, the clear region available to different drivers merely differs in scale and displacement across the visual field. These differences are mostly a result of the different relative head position of each driver, something that is in turn determined by their height and their individual adjustment of the seat for comfortable driving. Differences across drivers in Figure 6 are smaller than differences within drivers during natural driving, the latter being a consequence of changes in the position of the head forced by road characteristics. These within-driver differences are described next.
Variations in the Available Visual Field When Head Position Changes During Natural Driving
The absence of occlusive elements in the interior of the automobile would allow a driver to negotiate bends and turns with only eye movements and head rotations (as bicycle riders do), but the existing occlusive elements encroach into central vision in many circumstances. Hence, drivers need to tilt their head and move it sideways or forward to see around these elements. Figure 6 illustrates this characteristic by showing snapshots from a video clip recorded while subject 1 was driving the Citroën Xsara Picasso (the most obstructive automobile in our set) on a winding mountain road. In the picture on the left, the subject was driving on a straight stretch, and his available visual field is similar to that shown in the center panel in the center row of Figure 4 (for the same subject on the same automobile but with the vehicle parked); in the picture on the right, the subject was taking a sharp left bend on the same road, and he moved his head forward and tilted it to the right in an attempt to go around the obscuration caused by the large and split left A-pillar.
Note that relocation of the head allows clear central vision of the road just ahead of the automobile, but its effect is also that the distance between the eyes and the left A-pillar decreases. Hence, the obscuration caused by the A-pillar in near peripheral vision is wider (approximately 30° horizontally at eye height as determined by overlaying our measurement grid on the image) than it is when the driver only rotates his head (approximately 20°; see the left panel in the center row of Fig. 4).
Visual Fields in Urban Driving
Visual field occlusions caused by the driver's automobile threaten safe driving in the conditions of winding mountain roads, but these occlusions become a larger threat in the tight and packed conditions of urban spaces. The snapshot on the left of Figure 7 shows the visual information available to a driver who is about to complete a right turn at a narrow intersection, where only the stripes of a zebra pass on the pavement and the pole of a traffic light on the right serve as a warning sign that pedestrians may have the right of way. (Neither these nor other types of warning are available to the driver before or during the maneuver.) There is indeed a traffic light warning drivers to give way to pedestrians, but it is completely occluded by the automobile when the driver keeps a natural driving posture (see Fig. 7B) and becomes visible only when drivers lean forward and turn their head completely to the right (see Fig. 7C).
DISCUSSION AND CONCLUSION
The use of our technique reveals that vehicles differ in the way in which they restrict the driver's visual field. The largest and potentially more harmful occlusions for winding-road driving are caused by A-pillars at the driver's eye height (where these pillars are thicker and closer to the driver's eyes as a result of their raked design) and also by the interior rearview mirror and its attachment; in urban driving, the roof of the automobile also obstructs the view of traffic lights and signs when making sharp turns at narrow intersections. Video recordings during natural driving have further shown that these occlusive elements dangerously enter the line of gaze of the driver on many occasions.
Considering that a driver must have a clear view of a sufficiently long stretch of the road ahead as well as a clear view sideways so as to be able to see crossing pedestrians when making sharp turns at narrow urban intersections, the obscuration caused by elements in the interior of the automobile threatens safe driving. These elements (which are there for safety) produce visual defects that encroach into the 120° horizontal visual field that is generally deemed necessary for driving and which applicants for a driver's license are required to have.16,17
The method demonstrated here can be used in a number of areas. The most obvious one is in accident investigation to determine the extent to which occlusive elements in the automobile involved in the accident might have had a role in its occurrence. According to police files, there were 1782 accidents involving pedestrians in the city of Madrid, Spain, in the year 2003, yielding 29 dead and 1916 injured. A number of these accidents occurred at narrow intersections with traffic lights located where a driver can hardly see them, but the database does not include information (car make and model) that could be used to discern to what extent the occlusions considered in this article had a role.
The technique demonstrated here can also be used in a number of related applications, including the determination of the optimal location of traffic lights at urban intersections so that they are not occluded by the vehicle that the driver is on or the determination of the extent of a parking clearance region around crosswalks and intersections to facilitate the detection of crossing pedestrians (video recordings not shown here reveal that parked cars severely restrict the already poor visual field of a driver who is making this maneuver). This technique could also be used to suggest directions for improving the design of automobiles to minimize the obscuration that its elements necessarily produce (regarding the minimization of A-pillar obscuration, see a prototype at www.conceptlabvolvo.com/vcc/scc). Finally, it could be also used for the assessment of visual fields while the driver is glancing at in-car targets (e.g., the speedometer), like in the study of Summala.18
We are currently working on an improved apparatus that will use two cameras separated by the driver's interpupillary distance for additional precision in all these applications.
This work was supported by grant FIS–PI021829 from Instituto de Salud Carlos III (Ministerio de Sanidad y Consumo) and by grant BSO2001–1685 from Ministerio de Ciencia y Tecnología. The authors thank Mr. Jesús Mora de la Cruz (Coordinador General de Seguridad, Ayuntamiento de Madrid) for providing us with summary data on accidents involving pedestrians in the city of Madrid.
Miguel A. García-Pérez, PhD
Departamento de Metodología
Facultad de Psicología
Universidad Complutense, Campus de Somosaguas
28223 Madrid, Spain
1. Steel SE, Mackie SW, Walsh G. Visual field defects due to spectacle frames: their prediction and relationship to UK driving standards. Ophthal Physiol Opt 1996;16:95–100.
2. Roscoe SN, Hull JC. Cockpit visibility and contrail detection. Hum Factors 1982;24:659–72.
4. Cairney P, Catchpole J. Patterns of perceptual failures at intersections of arterial roads and local streets. In: Gale AG, Brown ID, Haslegrave CM, Taylor SP, eds. Vision in Vehicles—V. Amsterdam: Elsevier; 1996:87–94.
5. Herslund MB, Jorgensen NO. Looked-but-failed-to-see-errors in traffic. Accid Anal Prev 2003;35:885–91.
6. Reason JT, Manstead A, Stradling S, Baxter JS, Campbell K. Errors and violations on the roads: a real distinction? Ergonomics 1990;33:1315–32.
7. Aberg L, Rimmö PA. Dimensions of aberrant driver behaviour. Ergonomics 1998;41:39–56.
8. Summala H, Pasanen E, Rasanen M, Sievanen J. Bicycle accidents and drivers' visual search at left and right turns. Accid Anal Prev 1996;28:147–53.
9. Clarke DD, Ward PJ, Jones J. Processes and countermeasures in overtaking road accidents. Ergonomics 1999;42:846–67.
11. Hills BL. Vision, visibility, and perception in driving. Perception 1980;9:183–216.
12. Allen MJ Visual problems in the automobile. In: Allen MJ, Abrams BS, Ginsburg AP, Weintraub L, eds. Forensic Aspects of Vision and Highway Safety. Tucson, AZ: Lawyers & Judges Pub Co; 1996:111–47.
14. Vargas-Martín F. Disposable camera viewfinders as visual field expanders for tunnel vision. Ophthalmic Res 2004;36S1:206.
15. Bennett AG, Rabbetts RB. Bennett & Rabbetts' Clinical Visual Optics, 3rd ed. Boston: Butterworth-Heinemann; 1998.
17. Peli E, Peli D. Driving With Confidence: A Practical Guide to Driving With Low Vision. River Edge, NJ: World Scientific; 2002.
18. Summala H. Forced peripheral vision driving paradigm: evidence for the hypothesis that car drivers learn to keep in lane with peripheral vision. In: Gale AG, Brown ID, Haslegrave CM, Taylor SP, eds. Vision in Vehicles—VI. Amsterdam: Elsevier; 1998:51–60