visual target detection models for civil twilight and night driving conditions

Upload: nenad211

Post on 04-Feb-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    1/32

    2

    Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    Helmut T. Zwahlen, Ph.D.

    Russ Professor

    Human Factors and Ergonomics Laboratory

    Department of Industrial and Manufacturing Systems Engineering

    Ohio University, Athens Ohio 45701-2979

    (740) 593-1550

    Thomas Schnell, Ph.D.

    Assistant Professor

    Cognitive Human Factors Laboratory

    Department of Industrial Engineering

    The University of Iowa

    Iowa City, Iowa 52242-1527

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    2/32

    2

    ABSTRACT

    A luminance contrast based computer visibility model is discussed in this paper and compared with the

    civil twilight method which has recently been introduced. The civil twilight method attempts to predict the

    visibility of ordinary objects (reflectance 3% to 79%, average size) using only the headlamp illuminance

    at the target. It is suggested by the authors that the one-factor approach used by the civil twilight method

    is insufficient to satisfactorily address target visibility in the field. Developers of more advanced visibility

    models generally attempt to design their models based on the current state of the visibility research and

    with enough capability to obtain a reasonable degree of realism. The authors consider the level of the

    benchmark illuminance (3.2 lx) used in the civil twilight method to be too high, leading to very short

    detection distances for pedestrians (refer to literature review) under automobile headlamp illumination at

    night. The developers of the civil twilight method claim that the 3.2 lx visibility benchmark is based on

    systematic visual observations made by astronomers over a century ago. The use of the civil twilight

    method for pedestrian detection under automobile headlamp illumination at night is strongly discouraged

    by the authors of this paper, because the method may be misused by forensic experts if there is a

    need to produce arbitrarily short pedestrian detection distances, irrespective of factors including the

    clothing reflectance, contrast, pedestrian size, windshield transmittance, atmospheric transmissivity, etc.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    3/32

    3

    INTRODUCTION

    The detection of targets under low light conditions has been of great interest to the scientific visibility

    community in the recent past. A rather large body of research is available, providing the scientific

    framework for target detection and target visibility, both from a theoretical and empirical point of view.

    The list of visibility research that investigated human visual performance is too long to be completely

    covered within the scope of this paper. It is, however, generally accepted that Blackwell [1][2][3] was a

    major contributor of visual performance data forming the basis of our understanding of the capabilities

    and limitations of the human visual system. More or less elaborate visibility models have been

    developed in recent years based on such empirical visual performance data. Visibility models have

    become an invaluable tool for predicting the visibility of targets under a wide range of viewing conditions.

    Recent visibility models include the TARDEC model (army model) [4][5][6][7], the CIE (Committee

    Internationale de LEclairage) model [8], the PCDETECT model [9], Adrians model [10], and computer

    based visibility models developed by the authors [11][12][13][14]. These models are fairly elaborate and

    were generally designed based on recent visibility research. The aim generally is to provide a visibility

    model that is sufficiently precise (near true target value) and sufficiently accurate (small dispersion) in its

    predictions under selected (possibly wide) range of input parameters. Another approach to target

    visibility predictions was taken by Owens et al [15] and Andre and Owens [16] with their civil twilight

    method. On the surface, their approach appears to provide a simple, holistic benchmark method to

    address the issue of target visibility under automobile low-beam illumination at night. However, their

    method has only 1 input parameter, namely the illumination at a point ahead of an automobile at a

    selected height, on a plane normal to the illumination axis. It should be evident to readers who are

    familiar with the basics of human visual performance at low light levels, that a one factor approach is

    insufficient address target visibility in the field. It seems that the use of illumination at the target as a

    visibility benchmark is a poor choice anyway. A better choice would be to use the target luminance,

    which at least would implicitly account for target reflectance. That would, of course require exact

    knowledge of the target reflectance properties. The civil twilight method was deliberately kept simple by

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    4/32

    4

    Owens et al. [15] and Andre and Owens [16], so that the only instrument needed to conduct

    measurements in the field would be a illuminance meter.

    The members of the visibility research community should rightfully be concerned about the

    proliferation of general, holistic approaches such as the civil twilight method. Such a holistic approach to

    a specialized field of research seems to be an attempt to undermine and reverse recent advances in

    visibility modeling. Human visual performance is not a field that should be trivialized by one-size-fits all

    benchmarks such as the civil twilight method. In light of the vast scientific knowledge that was gathered

    to date by many visibility researchers, it is hard to understand why Andre and Owens [16] pose the

    general question of how much light do we need to see ? The answer is, it depends. It depends on the

    visual target characteristics, on the background characteristics, on characteristics of the light source, on

    environmental characteristics such as atmospheric transmissivity, ambient illumination, and on a

    multitude of observer characteristics including (but not limited to) age, luminance and color contrast

    sensitivity, adaptation level, probability of detection, arousal, expectancy and retinal eccentricity [17]. In

    dynamic settings, the list of independent variables may be significantly expanded. It should be noted that

    even the most advanced visibility model couldn't guarantee that a given observer would really be able to

    detect a target exactly as predicted by the model all the time. In real world settings there may be

    unknown physical factors such as dirt on the headlamps or the windshield, transient glare, or unknown

    cognitive factors such as focus of attention, distraction, workload, etc. that could potentially influence the

    visibility of a target. Although the influence of many of these factors can be modeled, it is not always

    sure that their presence is known in a specific case. While even the most advanced current visibility

    models are not able to explain all of the variability, it is fairly self evident that they provide far more

    precise (near true target value) and accurate (small dispersion) predictions for a specific visibility

    situation than a simple model such as the civil twilight method.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    5/32

    5

    Statement of the Problem

    Luminance contrast based driver visibility models represent an important cornerstone in the design of

    automobile lighting systems, the design of retro-reflective sheeting materials and pavement markings,

    the design of traffic signals, etc. Visibility models are also often used by forensics experts in litigation

    cases involving automobile accidents at night. There are a number of relatively advanced and higher

    fidelity visibility models [4][5][6][7][8][9][10][11][12][13][14] that are based on the recent and/or present

    state of the art in visibility research. These models consider many of the relevant factors of influence.

    Thanks to being packaged in computer software, most of these models can be used by specialists with

    relative ease.

    Owens at al. [15] and Andre and Owens [16] suggest a method by which almost anyone, non-

    experts and experts, could categorize a visual target seen under automobile headlamp illumination into

    the visible or not visible category, simply by using an illuminance meter. It is our opinion that visibility

    modeling should remain the domain of visibility researchers only. Non-experts should consult experts if

    they seek a simple answer to the difficult question of how much light do we need to see. Serious

    visibility experts know that the answer to that question depends on many issues, non-experts on the

    other hand may not be aware of this. The civil twilight method attempts to determine the detection

    distance of a target seen at a selected height under automobile headlamp illumination by using the

    amount of headlamp illuminance at the target, on a plane normal to the illumination axis as the only

    driving factor.

    In addition, the level of the benchmark illuminance (3.2 lx) chosen by Owens et al [15] is

    considered to be too high, leading to very short detection distances (refer to literature review) under

    automobile headlamp illumination at night. Also, the civil twilight method proposed by Owens et al. [15]

    does not consider two of the major independent variables, namely target size and reflectance. There is a

    need to discuss the advantages, disadvantages, and the structure of visual target detection models as

    they relate to dawn, dusk, and night driving conditions.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    6/32

    6

    REVIEW OF THE TECHNICAL LITERATURE

    Owens, Francis, and Leibowitz [15] proposed a simple, functional approach, the civil twilight method,

    that according to their claims, allows for quantification of the nighttime motorist visibility under

    automobile illumination. The reader should refer to [18] for an additional discussion of the civil twilight

    method [15].

    The civil twilight method proposed by Owens et al. [15] determines the visibility distance of a

    target seen (at a selected height) under automobile headlamp illumination by locating the distance

    ahead of the vehicle at which the illuminance provided by the headlamps equals the twilight illuminance

    ETD[lx] as indicated by Equation (1)

    TD = CD0.3

    (1)

    where TD is the twilight distance in feet, CD is the luminous intensity in candelas of the (Cyclops)

    headlamp in the direction of the target, and 0.3 is the twilight illumination in footcandles (3.2 lx). It should

    be noted, that Owens et al. [15] do not consider the two headlamps to be laterally separated but rather

    to be both located in the center of the vehicle (cyclops geometry). Owens et al. [15] only refer to the

    targets as being ordinary objects, implying that the target reflectance may vary from R=3% to R=79%.

    Simply claiming that a visibility benchmark method works for all ordinary objects is vague and

    misleading, and it appears that such a method makes very little use of the current state of the art in

    visibility research.

    Andre and Owens [16] state that they have tested the predictions of their civil twilight method in

    the field. They used an automobile (1990 Buick Skylark) with aligned headlamps and simply measured

    the illumination from the headlamps at specific grid-points ahead of the automobile at a height of 12,

    35, and 59. Andre and Owens [16] found that the distances at which an illuminance of 3.2 lx was

    measured agreed rather well with the twilight plateaus provided by their civil twilight method. It should be

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    7/32

    7

    noted that calling such measurements a test of predictions provided by a visibility model is highly

    misleading. Andre and Owens [16] simply established the iso-illuminance distances at selected grid

    points and vertical heights, nothing more. Their measurements prove nothing, whatsoever, about the

    visual performance of human observers under specific conditions. One may agree that such

    measurements are useful for checking the accuracy of physical, photometrical quantities. A meaningful

    validation, however, would have to include a wide range of observers whose task is to detect a wide

    range of targets under a wide range of conditions. Such an experiment would be bound to disclose the

    shortcomings of the civil twilight method because this method is unable to explain the large variation that

    would be found in such a field experiment.

    Other target detection refernces that are discussed in detail in [18] are Hazlett, and Allen [19],

    Shinar [20], Chrysler et al. [21], Blomberg et al. [22], Olson and Sivak [23], Strickland, Ward, and Allen

    [24], Austin, Klassen, and Vanstrum [25], Zwahlen and Schnell [26], Blackwell [1], Blackwell and

    Blackwell [2]. Target detection distances from some of the above references are tabulated in Table 1.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    8/32

    8

    Basic Issues Regarding Visibility Modeling

    In general, models of processes are often developed to demonstrate a concept, to facilitate the

    understanding of a concept or process and to describe, in an abstract form, a process the way we think

    it works. Sometimes, models are used to predict the behavior of a system without disturbing the system

    itself. While approximations are sometimes used in engineering, one should be aware that for such

    approximations to be useful, they must provide fairly accurate (small dispersion) and precise (near true

    target value) results, they must be validated, and they should be accepted by the majority of the

    scientists and engineers. The reader should note that the human performance part of the civil twilight

    method has not been validated in the field. Certainly, it would be relatively simple to come up with test

    cases involving the factors that are neglected by the civil twilight method to demonstrate that the civil

    twilight method may provide completely inaccurate detection distance predictions. Figure 1a illustrates a

    basic fact of model building. In order to increase the degree of realism obtained with a model, one has to

    increase the complexity and the completeness. It should be noted that Owens et al. [15] claim that the

    civil twilight method implicitly accounts for a number of factors by virtue of the generality of the civil

    twilight method. The civil twilight method is supposed to encompass the visibility of all targets with an

    imaginary average reflectance, an average size, an average background luminance, etc. It is, however,

    quite evident that the civil twilight method with its rigid 3.2 lx benchmark value is unable to explain the

    large variation of detection distances that are observed for different targets illuminated under different

    conditions, and observed by different observers. In model building, one aims at determining the main

    effects and as many interaction effects as possible, in order to obtain a high correlation between the

    predicted and observed values. Usually, the model with the highest correlation (R2) should be selected

    for use until a better model is found.

    The visibility research community is called upon to conduct the research needed to build even

    more complete models. There should be no doubt that todays state of the art visibility models far

    exceed the capability of the civil twilight method in explaining the variability found in the visual detection

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    9/32

    9

    of targets. By no means, however, should one expect a visibility model ever to be free of variation and to

    be able to provide precise pinpoint predictions of target visibility. It is well known that human visual

    performance follows psychometric functions such as the ones established by Blackwell [1][2]. There will

    always be a less than perfect fit (R2

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    10/32

    10

    computers it is fairly easy to obtain the results of target detection models almost instantaneously, thus

    further reducing the need for a simple approach such as the civil twilight method.

    The Civil Twilight Method, Old Science Taken Out of Context

    Owens et al. [15] and Andre and Owens [16] state that their 3.2 lx visibility benchmark is based on

    systematic visual observations made by astronomers over a century ago. The simple fact that some

    research is over 100 years old does not necessarily lend more credence to its adequacy, especially not

    if the research is taken out of context 100 years later. Astronomers back then and today are generally

    more concerned about the visibility of astronomical objects rather than target visibility under automobile

    illumination. During WWII, the office of scientific research and development established the Tiffany

    foundation to further the state of the knowledge in visibility research. A large scale study funded by the

    Tiffany foundation was conducted by Blackwell [1], based on the need of the US Navy to learn more

    about the threshold contrast of the human eye. This large and relatively expensive study would not have

    been commissioned, if the designers and visibility scientists at the time had felt that they already had a

    handle on target visibility with the civil twilight observation data provided by astronomers. If these

    researchers already felt they needed more adequate data to better understand target visibility it would

    appear that more recent data provided by Blackwell [1][2] and other visibility researchers supersedes

    observations made by astronomers over 100 years ago. This is especially true since modern

    psychophysical experiments were conducted under controlled conditions, allowing for the isolation of

    main effects and interaction effects. The human factors and Ergonomics Laboratory at Ohio University,

    Athens, Ohio, used extensive visibility field research data to allow the use of laboratory visibility

    research data [1][2] in models that predict target visibility under specific real-world conditions (calibration

    of model). No psychophysical target detection field research was ever published to determine the validity

    of the civil twilight method.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    11/32

    11

    LUMINANCE AND ILLUMINANCE MESUREMENTS DURING CIVIL TWILIGHT ON AN OVERCAST

    EVENING AND ON A CLEAR EVENING

    It was hypothesized by the authors of this paper that the ambient illuminance and target luminances

    during civil twilight are highly dependent on weather conditions and on the direction in which the

    measurements are performed. The measurements reported in this section should provide the reader

    with an idea of the range of available ambient natural illuminance and its directionality during the civil

    twilight period on an overcast evening and on a clear evening. In addition, luminances were measured

    on a gray target and on a white target to obtain the range of luminances as a function of time and

    direction during the civil twilight period. It should be noted, that Owens et al. [15] chose an illuminance of

    3.2 lx, obtained at the end of the civil twilight on a clear day, as their benchmark illuminance ETDin the

    civil twilight method. An overcast evening was chosen for part of the measurements reported herein to

    demonstrate that the twilight illuminances can be considerably lower than the value of 3.2 lx reported by

    Owens et al. [15] under such conditions. The clear evening measurements were conducted to confirm

    that one would obtain twilight illumination values that are considerably higher than the ones measured

    during an overcast evening.

    Twilight is defined as the transition period from day to night (or night to day) when the sky is not

    completely dark. It is technically defined as the period of time beginning (or ending) when the center of

    the (refracted) sun is lower than a given elevation. Civil twilight corresponds to the sun being between 0

    and 6 below the horizon, nautical twilight to 6 and 12, and astronomical twilight to 12 and 18.

    Measurement Site, Setup, and Method

    The measurements were conducted on the beach parking lot of the Strouds Run State Park outside

    Athens, Ohio (Longitude:82o

    , 06', 20" W, Latitude:39o

    , 20', 14"N) during the civil twilight on February 3,

    1998 (overcast evening) and on March 26, 1998 (clear evening). The measurement cycle was repeated

    from 30 minutes prior to the beginning of the evening civil twilight to 30 minutes after the end of the

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    12/32

    12

    evening civil twilight. Care was taken to ensure that virtually no alternative (other than natural) sources

    of illumination were present. The gray target had a reflectance of R=0.108 and on the white target had a

    reflectance of R=0.92. Both targets were made of posterboard 0.6m x 0.6m. The gray target was spray

    painted with a blend of Krylon Satin Black spray paint and Krylon Satin White spray paint. The white

    target was left untreated. For the luminance measurements, the targets (see Figure 2) were positioned

    in 8 directions: North, Northeast, East, Southeast, South, Southwest, West, and Northwest. The

    measurements consisted of repeated cycles of one full revolution of luminance measurements (white

    and gray target in each position as indicated in Figure 2) followed by one full revolution of illuminance

    measurements including one measurement taken skywards (straight up). The targets were removed for

    illuminance measurements. At the end of the illuminance measurement cycle, the Pritchard 1980A was

    tilted 90oup to obtain the sky illuminance.

    Results of the Luminance and Illuminance Measurements

    Figure 3, Figure 4, and Figure 5 summarize the luminance and illuminance values measured on the

    overcast evening (February 3, 1998) and on a clear evening (March 26, 1998) at the Strouds run parking

    lot site. The measured luminance values (overcast condition) for the white target with a reflectance of

    R=0.92 are shown in Figure 3a. In spite of the diffuse overcast sky conditions, a small directionality

    effect was found as evidenced by the spread of the luminance curves in Figure 3a. At 6:19pm (end of

    the overcast evening civil twilight on February 3, 1998), the white target provided an average luminance

    of 0.012 cd/m2. Figure 3b shows the measured luminance values (overcast evening) for the gray target

    with a reflectance of R=0.108. At 6:19pm (end of the evening civil twilight), the gray target provided an

    average luminance of 0.0012 cd/m2.

    Figure 4a shows the luminance values that were measured during the evening civil twilight of

    March 26, 1998 (clear evening) on the white target. Again, a substantial directionality effect is observed

    and the average luminance of the white target at the end of the civil twilight is 0.52 cd/m2. The clear

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    13/32

    13

    evening civil twilight luminance values of the gray target are shown in Figure 4b. The average luminance

    of the gray target at the end of the civil twilight is 0.042 cd/m2.

    The ambient illuminances that were measured on February 3, 1998 (overcast evening) at the

    beach parking lot of the Strouds Run State park are shown in Figure 5a. Again, a directionality effect

    was found. At 6:19pm (end of the evening civil twilight) the average illuminance measured was 0.08 lx.

    Figure 5b shows the ambient illuminance [lx] as a function of time with direction as parameter, during the

    evening civil twilight of March 26, 1998 (clear sky). As in the overcast evening data (Figure 5a), there

    was a fairly large directionality effect. The average illuminance over all measured directions at the end of

    the civil twilight was 1.7 lx, which is still 46.9% lower than the 3.2 lx stated by Owens et al. [15].

    It appears that the series of measurements conducted under clear sky conditions come much

    closer to the 3.2 lx used by the civil twilight method. Benchmarks should not be based on highly variable

    phenomena but rather on highly reproducible phenomena. For the specific case of target visibility

    modeling the authors recommend against the use of benchmarks, whatsoever. However, the authors

    recommend the use of more precise, calibrated visibility models that produce unbiased estimates and

    account for as much of the variation as the current state of research permits. Looking at Figure 5 one

    can see that each illuminance line is a well behaved, monotonously decreasing function over time.

    Therefore, selecting an end of civil twilight illuminance of 3.2 lx is definitely arbitrary since the measured

    illuminances at the end of the civil twilight ranged from 0.06 lx (overcast sky) to 3.2 lx (clear sky). The

    luminances measured on the targets at the end of the civil twilight period ranged from 0.0009 cd/m2

    (gray target, overcast sky) to 1.2 cd/m2 (white target, clear sky). The measured luminance range

    practically covers the entire mesopic range. The practice of using the end of civil twilight illuminance as

    benchmark is highly questionable, since the photometric conditions associated with the end of civil

    twilight appear to span the entire mesopic range.

    Figure 6a shows twilight distance plateaus for an automobile using Ford Taurus low-beam

    headlamps and a target (size and reflectance are not considered by twilight distance method) with an

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    14/32

    14

    assumed vertical target center at 0.53m above ground. The plateaus were constructed for various

    threshold twilight illuminances. Using a twilight illuminance of 3.2 lx, it can be seen that the twilight

    distance is about 43m, if the target is placed 1.5m to the right of the longitudinal vehicle axis. A twilight

    illuminance of 1.6 lx allows the twilight distance to increase to 60m, if the target is located about 2.2m to

    the right of the longitudinal vehicle axis. Decreasing the twilight illuminance to 0.8 lx allows the twilight

    distance to increase to 87m, if the target is located about 3.5m to the right of the longitudinal vehicle

    axis. A twilight distance of over 100m can be obtained if the twilight illuminance is assumed to be only

    0.4 lx and the target is located 2.5m to the right of the longitudinal vehicle axis. Figure 6a basically

    illustrates that an arbitrary selection of the twilight illuminance will result in an equally arbitrary twilight

    distance. The illuminance measurements have clearly shown, that an illuminance of 3.2 lx seems quite

    high. Choosing a twilight illumination of 3.2 lx will consequently lead to relatively short twilight distances

    as indicated by Figure 6a. Again, it seems that the twilight distance method is a completely inadequate

    approach to solving a complex visual detection modeling problem.

    Comparison of Detection Distances Obtained Using the Civil Twilight Method With Detection

    Distances Obtained Using a State of the Art Computer Model

    It was previously mentioned that the civil twilight method neglects a number of factors that are known to

    significantly affect the visibility of a target. This section demonstrates the effect of the angular target size,

    the target reflectance, and the driver age upon the detection distance of selected targets. The obtained

    detection distances are compared to the corresponding twilight distances.

    A proprietary computer based visibility model was used by the authors to calculate the detection

    distances for the selected targets. This computer visibility model is based on the Blackwell human

    threshold contrast database [1]. The age function [11] used in the computer visibility model is based on

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    15/32

    15

    data provided by Blackwell for 156 normal observers of various ages [2]. The headlamps, the observer,

    and the targets are accurately accounted for in 3D space.

    Figure 6b illustrates the setup used for the calculations of the detection distance of diffuse targets

    located on the right road shoulder. The center of the targets are located 0.53 m above the road surface.

    To demonstrate the effect of the target size it is assumed that the target size is increased along the

    vertical dimension from 0.13m up to 1.06m. Two target reflectances R1=0.3 and R2=0.1 are used to

    demonstrate the effect of target reflectance. The driver eye location and the headlamp location are valid

    for a 50 percentile adult in an average car [36]. In computing the visibility of a given target, the

    proprietary computer based visibility model first determines the luminance of the background along the

    longitudinal target axis by using a road surface reflectance matrix (used asphalt in this case). It should

    be noted, that only the luminance of the pavement was used as a target background. It was assumed,

    that the target did not extend into the horizon sky. Candlepower matrices are used for each headlamp

    separately to determine the illuminance at the target. Then the computer model determines the

    luminance of the target along the longitudinal target axis, using the reflectance of the diffuse target. With

    the target luminance LT and background luminance LB, the computer model determines the actual

    contrast along the longitudinal target axis using Equation (2).

    CACT =LT L B

    L B(2)

    Then, the computer visibility model compares the actual contrast with the threshold contrast CTH

    determined from Blackwells human threshold contrast database [1]. The reader is referred to [14] for a

    detailed description of the interpolation method used to determine CTH as a function of the visual angle

    subtended by the target and the available background Luminance LB. In a final step, the computer

    visibility model applies a number of contrast multipliers to account for the observer age [2][11] (25 years

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    16/32

    16

    and 65 years used in present paper), probability of detection (P=99.93%1used in this paper), exposure

    time [37] (0.65 seconds used in this paper), and a field factor of 10 to account for the difference between

    Blackwells trained observers in the laboratory [1] and normal observers driving in an automobile. The

    detection distance is given by the longitudinal location at which the actual contrast CACT

    equals the

    adjusted threshold contrast CTH.

    Figure 7 clearly illustrates the effect of Target Size and Target Reflectance on the detection

    distance using the proprietary computer based visibility model. Figure 7a shows that young observers

    (25 years) can detect the dark achromatic targets with a reflectance of R=0.1 at a detection distance of

    30.0m, 36.0m, 39.3m, and 41.5m for a target height of 0.13m, 0.26m, 0.53m, and 1.06m, respectively.

    All of these distances are shorter than the twilight distance (headlamp illumination equals 3.2 lx, 0.53m

    above ground) of 53m. Figure 7b shows that young observers (25 years) can detect the light achromatic

    targets with a reflectance of R=0.3 at a distance of 62.5m, 77.5m, 87.0m, and 93.0m for a target height

    of 0.13m, 0.26m, 0.53m, and 1.06m, respectively. All of these distances are longer than the twilight

    distance (headlamp illumination equals 3.2 lx, 0.53m above ground) of 53m. Old observers (Figure 7b)

    detect the 1.06m high target with a reflectance of R=0.3 at a distance of only 77.5m which equates to a

    loss in the visibility distance of 16.6% over the young observers. Figure 7 provides evidence that the civil

    twilight distance method is not only unable to account for target size, but also fails to predict the severe

    reduction in detection distance when using dark, diffuse target materials. Also, the civil twilight distance

    method does not consider the factors of contrast and observer age, to name a few.

    1Probability of Detection of 99.93% (3.2) means that in 9,993 out of 10,000 observations the target is

    detected

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    17/32

    17

    Summary, Discussion, and Conclusion

    This paper investigated and identified the shortcomings of the civil twilight method proposed by Owens

    et al. [15] and Andre and Owens [16] as a simple functional approach to determine the visibility distance

    of an ordinary object under automobile illumination at night. The civil twilight method attempts to

    determine the detection distance of a target seen under automobile headlamp illumination at night by

    using the amount of illuminance provided by the headlamps as the only benchmark factor. The civil

    twilight method completely neglects target size and reflectance. In other words, for the civil twilight

    method it does not matter whether the target is a dark clad pedestrian or a large light colored farm

    animal. Other factors that are neglected by the civil twilight method, in spite of scientific evidence

    indicating them to be important for visbility modeling are driver age, windshield transmittance, glare,

    atmospheric transmissivity, and exposure time, just to name a few.

    A series of luminance and illuminance measurements were conducted during the evening civil

    twilight of February 3, 1998 (overcast evening) and March 26, 1998 (clear evening). A strong

    dependency of the measured illuminance on the prevailing weather condition and a directionality effect

    were found. The measured illuminance of 0.08 lx at the end of the civil twilight during the overcast

    evening is 40 times lower than the 3.2 lx cited by Owens et al. [15]. The measured illuminance of 1.7 lx

    at the end of the civil twilight during the clear evening is still 1.9 times lower than the 3.2 lx cited by

    Owens et al. [15].

    It is the opinion of the authors of this paper that the civil twilight method as a whole appears to be

    inadequate and insensitive, and should not be used at all. Owens et al. [15] feel that the value of their

    method lies in its simplicity rather than in its accuracy. It is the opinion of the authors of this paper that

    proposing a simple, non-validated approach such as the civil twilight method to solve target detection

    problems under automobile headlamp illumination at night, is highly questionable and the results of the

    civil twilight method may be misleading at best. There really is no need for a simple target detection

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    18/32

    18

    algorithm like the civil twilight method, since advanced target detection/visibility/legibility models are

    available as computer programs that can be used by specialists. At this point in time, no claim should be

    made that the civil twilight method is an accepted, validated, adequate, scientific method to determine

    the visibility of targets.

    The use of the civil twilight method for pedestrian detection under automobile headlamp

    illumination at night is strongly discouraged by the authors of this paper, because the method may be

    misused by forensics experts if there is a need to produce arbitrarily short pedestrian detection

    distances, irrespective of factors including the clothing reflectance, contrast, pedestrian size, windshield

    transmittance, atmospheric transmissivity, etc. Advanced visibility models that are calibrated with field

    data and which consider a multitude of factors provide predictions that are more precise (near true target

    value) and more accurate (small dispersion) than predictions made by the civil twilight method.

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    19/32

    19

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    20/32

    20

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    21/32

    21

    0.001

    0.01

    0.1

    1

    10

    100

    5:30 5:35 5:40 5:45 5:50 5:55 6:00 6:05 6:10 6:15 6:20 6:25 6:30 6:35 6:40 6:45

    Time Begin

    Luminance[cd/m^2]

    SE

    S

    SW

    W

    NW

    N

    NE

    E

    Sunset: 5:52pm

    End of Civil Twilight : 6:19pm

    Luminance of a white target taken in

    different directions, 100% overcast.

    Strouds Run State Park, Athens, Ohio.

    Lat: 39:18:58N, Long: 82:05:42W

    Civil Twilight begins: 5:52pm, ends: 6:19pm.

    a. White Target, R=0.92

    0.0001

    0.001

    0.01

    0.1

    1

    10

    5:30 5:35 5:40 5:45 5:50 5:55 6:00 6:05 6:10 6:15 6:20 6:25 6:30 6:35 6:40 6:45

    Time Begin

    Luminance[cd/m

    ^2]

    SE

    S

    SW

    W

    NW

    N

    NE

    ESunset: 5:52pm

    End of Civil Twilight: 6:19pm

    Luminance of a gray target taken in

    different directions, 100% overcast.

    Strouds Run State Park, Athens, Ohio.

    Lat: 39:18:58N, Long: 82:05:42W

    Civil Twilight begins: 5:52pm, ends: 6:19pm.

    b. Gray Target, R=0.108

    Luminance Measurements Conducted on the Beach Parking Lot of the Strouds Run State Park Outside Athens, Ohio(Longitude:82

    o06 20" W, Latitude:39

    o20 14"N) During the Evening Civil Twilight on February 3, 1998

    Figure 3. Overcast Evening: Target Luminance [cd/m2] as a Function of Time with Direction as

    Parameter, During Evening Civil Twilight

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    22/32

    22

    0.0001

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    6:15 6:20 6:25 6:30 6:35 6:40 6:45 6:50 6:55 7:00 7:05 7:10 7:15 7:20 7:25 7:30 7:35 7:40

    Time at Beginning of Measurements

    Luminance[cd/m^2]

    SE White

    S White

    SW White

    W White

    NW White

    N White

    NE White

    E WhiteBeginning of Civil Tw ilight, 6:46

    End of Civil Tw ilight, 7:13

    Luminance of w hite square target in

    diff erent directions during Civil Tw ilight.

    26 March 1998. Clear sky. Strouds Run, Athens, OH.

    Latitude: 39:18:58N, Longitude: 82:05:42W

    a. White Target, R=0.92

    0.0001

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    6:15 6:20 6:25 6:30 6:35 6:40 6:45 6:50 6:55 7:00 7:05 7:10 7:15 7:20 7:25 7:30 7:35 7:40

    Time at Beginning of Measurements

    Luminance

    [cd/m^2]

    SE Gray

    S Gray

    SW Gray

    W Gray

    NW Gray

    N Gray

    NE Gray

    E GrayBeginning of Civil Tw ilight, 6:46

    End of Civil Tw ilight, 7:13

    Luminance of gray square target in diff erent

    directions dur ing Civil Twilight.

    26 March 1998. Clear sky. Strouds Run, Athens, OH.

    Latitude: 39:18:58N, Longitude: 82:05:42W

    b. Gray Target, R=0.108Luminance Measurements Conducted on the Beach Parking Lot of the Strouds Run State Park Outside Athens, Ohio(Longitude:82

    o06 20" W, Latitude:39

    o20 14"N) During the Evening Civil Twilight on March 26, 1998

    Figure 4. Clear Evening: Target Luminance [cd/m2] as a Function of Time with Direction as Parameter,

    During Evening Civil Twilight

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    23/32

    23

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    5:40 5:45 5:50 5:55 6:00 6:05 6:10 6:15 6:20 6:25 6:30 6:35 6:40 6:45

    Time Begin

    Illuminance[lux]

    SE

    S

    SW

    W

    NWN

    NE

    E

    Up

    3.2lx

    Illuminance taken in different directions, 100% overcast.

    Strouds Run State Park, Athens, Ohio.

    Lat: 39:18:58N, Long: 82:05:42W

    Civil Twilight begins: 5:52pm, ends: 6:19pm.

    3.2lx at 6:01pm

    when

    measuring NE

    3.2lx at 6:04pm

    when

    measuring up

    0.08lx at end

    of

    civil twilight

    a. Overcast Evening, February 3, 1998

    0.001

    0.01

    0.1

    1

    10

    100

    1000

    6:35 6:40 6:45 6:50 6:55 7:00 7:05 7:10 7:15 7:20 7:25 7:30 7:35 7:40 7:45 7:50

    Time at Beginning of Measurements

    Illuminanc

    e[lux]

    SE

    S

    SW

    W

    NW

    N

    NE

    E

    Up

    3.2 lx

    Illuminance in different directions during Civil Twilight.

    26 March 1998. Clear sky. Strouds Run, Athens, OH.

    Latitude: 39:18:58N Longitude: 82:05:42W

    Civil twilight begins 6:46pm, ends 7:13pm

    Stated value of illuminance

    at the end of Civil Twilight, 3.2 lux.

    3.2lx at 7:06pmwhen

    measuring

    East

    3.2lx at 7:13pm

    when

    measuring

    West

    b. Clear Evening, March 26, 1998Illuminance Measurements Conducted on the Beach Parking Lot of the Strouds Run State Park Outside Athens, Ohio(Longitude:82o06 20" W, Latitude:39o20 14"N) During the Evening Civil Twilight on February 3, 1998

    Figure 5. Illuminance [lx] as a Function of Time with Direction as Parameter, During Evening Civil

    Twilight

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    24/32

    24

    a. Twilight Distance Plateaus for Ford Taurus Low-Beam Headlamps, Assumed Vertical Target Centeris 0.53m above Ground

    Area enclosed by each plateau receives an illuminance at or above the indicated value in lux

    =Origin

    Driver

    LongitudinalOffset2.05m

    Longitudinal Dista nce to Target

    DriverLateral

    Offset0.32m

    Target Heightto Center 0.53m

    HeadlampHeight0.607m

    DriverEye Height

    1.16m

    Lane Width3.65m(12ft)

    1/2 Lane Width

    1.82m(6ft)

    1/2 Lane Width

    1.82m(6ft)

    Headlamp

    Separation1.11m(4ft)

    Vertical TargetDimension

    1.06m0.53m0.26m

    0.13m

    Diffuse TargetLocated on RightRoad Shoulder

    Reflectance10% and 30%Width 0.13m

    b. Setup used for the Calculation of the Detection Distance of Diffuse, Achromatic Targets

    Figure 6. Iso Twilight Distance Plateaus for Various Twilight Illuminations and Setup Used in Detection

    Distance Calculations

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    25/32

    25

    0.001

    0.01

    0.1

    1

    10

    0 5 10 15 20 25 30 35 40 45 50 55 60

    Distance [m]

    Contrast

    Cth for 1.06m high targetCth for 0.53m high targetCth for 0.26m high targetCth for 0.13m high targetCact

    Object with center located along right edge line (1.82m from car

    center) and 0.53 m above ground.

    Reflectance = 10%, H6054 low beams, Young Observer 25 years

    Probability of Detection P=99.93%

    Exposure Time 0.65seconds (85th percentile eye fixation duration)

    Field Factor = 10

    Twilight Distance

    = 53m

    (3.2 lux at target)

    30.0m 36.0m

    39.3m

    41.5m

    a. Reflectance R=0.1

    0.01

    0.1

    1

    10

    100

    0 10 20 30 40 50 60 70 80 90 100 110 120

    Distance [m]

    Contra

    st

    Cth for 1.06m high target, 65 yearsCth for 1.06m high target, 25 yearsCth for 0.53m high target, 25 yearsCth for 0.26m high target, 25 yearsCth for 0.13m high target, 25 yearsCact

    Twilight Distance

    = 53m

    (3.2 lux at target)

    Object with center located along right edge line (1.82m from car center)

    and 0.53 m above ground.

    Reflectance = 30%, H6054 low beams, Probability of Detection 99.93%

    Exposure Time 0.65seconds (85th percentile eye fixation duration)

    Field Factor = 10

    62.5m 77.5m 87.0m

    93.0m

    b. Reflectance of R=0.3

    Figure 7. Actual Contrast and Threshold Contrast as a Function of Longitudinal Distance for Diffuse,

    Achromatic Targets Ranging in Height from 0.13m to 1.06m, Target Center 0.53m above

    Ground, As Seen Against Used Asphalt, Reflectance R=0.1 and R=0.3

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    26/32

    26

    Table 1. Comparison of Average Target Detection Distances Reported by Various Researchers

    Note: Entries are sorted by increasing detection distance, entries shown in white on black indicate oncoming car glareconditions. NV=Not Visible

    Researcher(s) Method Observer Age Beam Type Target TypeTarget

    Reflectance/AppearanceTarget Size

    Average

    Detection

    Distance [m]Owens et al. Twilight Distance Not considered Europe low beams Any target, 43.3" above ground Not considered Not considered NV

    Olson and Sivak Field, Approaching Glare source young Low beams Human figure on Left Dark Adult human 18

    Olson and Sivak Field, Approaching Glare source young Low beams Human figure on right Dark Adult human 24Owens et al. Twilight Distance Not considered US low beams Any target, 43.3" above ground Not considered Not considered 29

    Leibowitz and Owens not stated Not considered Low beams Pedestrian in dark clothing Not specified Not specified 34

    Leibowitz and Owens not stated Not considered High beams Pedestrian in white clothing Not specified Not specified 34Hazlett and Allen Field, Blood Alcohol Content = 0 Young Low beams Simulated pedestrian Black cloth, 9% .3m x .3m x 1.2m 34

    Olson and Sivak Field, Approaching Glare source young Low beams Human figure on Left Light Adult human 37

    Olson and Sivak Field, Approaching Glare source young High beams Human figure on right Dark Adult human 37Olson and Sivak Field, Approaching Glare source young High beams Human figure on Left Dark Adult human 37

    Hazlett and Allen Field, Blood Alcohol Content = 0 Young Low beams Simulated pedestrian Gray cloth, 16% .3m x .3m x 1.2m 37

    Owens et al. Twilight Distance Not considered Europe low beams Any target, 27" above ground Not considered Not considered 44

    Leibowitz and Owens not stated Not considered High beams Pedestrian in dark clothing Not specified Not specified 49Olson and Sivak Field, Approaching Glare source young Low beams Human figure on right Light Adult human 49

    Owens et al. Twilight Distance Not considered US low beams Any target, 27" above ground Not considered Not considered 55

    Chrysler Field, rural test track 53-75 Low beams Small road hazard 31.5% Reflectance .17m x .33m 73Olson and Sivak Field, Approaching Glare source young High beams Human figure on right Light Adult human 73

    Olson and Sivak Field, Approaching Glare source young High beams Human figure on Left Light Adult human 73

    Owens et al. Twilight Distance Not considered US low beams Any target on ground level Not considered Not considered 86Chrysler Field, rural test track 53-75 Low beams Child mannequin 37.8% Reflectance 1.06m tall 88

    Owens et al. Twilight Distance Not considered Europe low beams Any target on ground level Not considered Not considered 91

    Bloomberg et al Field, dark test course Not specified Low beams Child, jacket Gray Garmet size 8 94

    Chrysler Field, rural test track 19-25 Low beams Small road hazard 31.5% Reflectance .17m x .33m 96

    Shinar Field, rural roads, target not expected 20-58 Low beams Pedestrian Khaki, 5% Reflectance Adult human 101Shinar Field, rural roads, target not expected 20-58 Low beams Pedestrian Khaki, 70% Reflectance Adult human 105

    Bloomberg et al Field, dark test course Not specified Low beams Adult, Coverall Gray Garmet size , large 110Bloomberg et al Field, dark test course Not specified Low beams Child jacket White Garmet size 8 114

    Owens et al. Twilight Distance Not considered US high beams Any target, 43.3" above ground Not considered Not considered 125

    Chrysler Field, rural test track 19-25 Low beams Child mannequin 37.8% Reflectance 1.06m tall 125Owens et al. Twilight Distance Not considered US high beams Any target, 27" above ground Not considered Not considered 137

    Bloomberg et al Field, dark test course Not specified Low beams Human adult figure target Gray Not specified 143

    Owens et al. Twilight Distance Not considered US high beams Any target on ground level Not considered Not considered 143

    Owens et al. Twilight Distance Not considered Europe High beams Any target on ground level Not considered Not considered 146Owens et al. Twilight Distance Not considered Europe High beams Any target, 27" above ground Not considered Not considered 146

    Owens et al. Twilight Distance Not considered Europe High beams Any target, 43.3" above ground Not considered Not considered 146

    Hazlett and Allen Field, Blood Alcohol Content = 0 Young Low beams Simulated pedestrian White cloth, 75% .3m x .3m x 1.2m 149

    Shinar

    Field, rural roads, target expected within2km, laterally located either on center, left

    or right 20-58 Low beams Pedestrian Khaki, 5% Reflectance Adult human 150

    Shinar

    Field, rural roads, target expected at exactlongitudinal distance, laterally located

    either on center, left or right 20-58 Low beams Pedestrian Khaki, 5% Reflectance Adult human 152

    Shinar

    Field, rural roads, target walking away

    from static car 20-58 Low beams Pedestrian Khaki, 70% Reflectance Adult human 160

    Shinar

    Field, rural roads, target walking away

    from static car 20-58 Low beams Pedestrian Khaki, 5% Reflectance Adult human 165

    Shinar

    Field, rural roads, target expected within

    2km, laterally located either on center, leftor right 20-58 Low beams Pedestrian Khaki, 70% Reflectance Adult human 175

    Bloomberg et al Field, dark test course Not specified Low beams Adult, Jacket White Garmet size 44 178

    Shinar

    Field, rural roads, target expected at exact

    longitudinal distance, laterally located

    either on center, left or right 20-58 Low beams Pedestrian Khaki, 70% Reflectance Adult human 185Bloomberg et al Field, dark test course Not specified Low beams Adult, Trousers White Garmet size , large 219

    Bloomberg et al Field, dark test course Not specified Low beams Human adult figure target White Not specified 257

    Bloomberg et al Field, dark test course Not specified Low beams Adult, Coverall White Garmet size , large 275

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    27/32

    27

    Table 2. Comparison of visibility Model Components

    Component present in Visibility Method

    Civil

    TwilightMethod

    Ohio

    University

    Visibility

    Model

    Component present in Visibility

    Method

    Civil

    TwilightMethod

    Ohio

    University

    Visibility

    Model

    Candlepower beam pattern yes yes Observer threshold contrast no yes

    individually defined with 6 dof no yes Threshold illumination at the eye no

    yes, based on

    Blackwell

    1946 data

    Observer location with 6 dof no yes Adaptation of observer no

    yes, based on

    Blackwell

    1946 data

    Windshield transmission no yes Age-background luminance interaction no

    yes, based on

    data from

    Blackwell

    1946 and 1971

    Atmosperic transmissivity no yes Target eye fixation duration no

    yes, Ohio

    University eye

    scanning

    research

    Above horizon scene background

    luminanceno yes Probability of detection no yes

    Road surface luminance (for pavement

    markings)no yes Disability glare no yes

    4 dimensional coefficient of retro-

    reflection matrices for micro-prismatic

    retro-reflectors

    no yes Contrast polarity no yes

    2 dimensional coefficient of retro-

    reflection matrices for beaded materialsno yes Effects of color contrast no no

    2 dimensional coefficient of retro-reflectance matrices for pavement

    markings and road surfaces

    no yesEffects of non-uniform background

    luminanceno no

    Reflectance of diffuse reflectors no yes Observer attention/arousal no no

    Target location and orientation in 6 dofno, 3 dof

    onlyyes

    Calibration with target and task

    specific psychometric visibility field

    data

    no yes

    Target size no yesValidation with target and task specific

    psychometric visibility field datano yes

    Model usable for legibility of text no yes

    Peripheral detection no noObserver age no yes

    dof = degrees of freedom

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    28/32

    28

    REFERENCES

    [1] Blackwell H.R., Contrast Thresholds of the Human Eye, Journal of the Optical Society of America,

    36(11), 1946

    [2] Blackwell O.M., Blackwell H.R., Visual Performance Data for 156 Normal Observers of Various Ages.

    Journal of the Illuminating Engineering Society, 1(1):pp 3-13, 1971.

    [3] Blackwell H.R., Specification of Interior Illumination Levels, Illuminating Engineering, 54, pp. 317-53,

    1959

    [4] Gerhart G., Meitzler T., Sohn E., Verification and Validation Status of the TARDEC Visual Model, In

    Proceedings of the Ground Target Modeling and Validation Conference, Houghton, MI, 1995

    [5] Lindquist G.H., Witus G., Cook T.H., Freeling J.R., Target Discrimination Using Computational Vision

    Human Perception Models, In Proceedings of the 8thInternational Symposium of the SPIE, Orlando,

    FL, 1994

    [6] Meitzler T., Bryk D., Gerhart G., Karlsen R., Sohn E., Witus G., Anneberg L., Evaluation of the

    Conspicuity of Civilian Vehicles Utilizing a Dual Purpose Visual Perception Laboratory [VPL] and

    Visual Perception Model [VPL], In Proceedings of the 7th Annual Ground Target Modeling and

    Validation Conference, Houghton, MI, 1996

    [7] Meitzler T., Gerhart G., Sohn E., Witus G., Heckmann T., Cusmano E., Adapting an Army Vision

    Model for Measuring Armored-Vehicle Camouflage to Evaluating the Conspicuity of Civilian

    Vehicles, In Proceedings of the 5th Annual Ground Target Modeling and Validation Conference,

    Houghton, MI, 1994

    [8] Commision Internationale de L Eclairage, An Analythic Model For Describing The Influence Of

    Lighting Parameters Upon Visual Performance, Publication CIE No. 19/2.1 (TC-3.1), Paris, France,

    1989

    [9] Farber E., C. Matle, PCDETECT: A Revisd Version of the DETECT Seeing Distance Model,

    Transportation Research Record 1213, Transportation Research Board, National Research Council,

    Washington, D.C. 1989, pp. 11-20

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    29/32

    29

    [10] Adrian W., Visibility of Targets, Transportation Research Record 1247, Transportation Research

    Board, National Research Council, Washington, D.C. 1989, pp. 39-45

    [11] Zwahlen, Helmut T. and Schnell, Thomas, "A Combined Age-Background Luminance Contrast

    Multiplication Function to Adjust the Human Contrast Threshold More Accurately in Visibility and

    Legibility Evaluations," paper presented at the PAL-Progress in Automobile Lighting-Symposium,

    September 26-27, 1995, Technical University, Darmstadt, Germany. Published in PAL Symposium

    Proceedings, Technische Hochschule Darmstadt, 1995, pp. 240-247.

    [12] Schnell T., Zwahlen H.T., The Development of CARVE, Computer Aided Road-Marking Visibility

    Evaluator, A PC-Based Pavement Marking Visibility Evaluation Software Package, Paper presented

    at the TRB A3A04 Visibility Committee Midyear Symposium on Night Visibility and Driver Behavior,

    Iowa City, IA, April 28-30, 1996

    [13] Schnell, Thomas and Zwahlen, Helmut T., "Predicting the Visibility of Pavement Markings with

    CARVE (Computer-Aided Road-Marking Visibility Evaluator)," paper presented at the Ohio

    Transportation Engineering Conference, The Ohio State University, Columbus, Ohio, November 19-

    20, 1996.

    [14] Schnell T., Zwahlen H.T., Retroreflective Materials, Visibility Calculations for Prismatic or Micro-

    Prismatic Materials, Paper presented at the Transportation Research Board 77th Annual Meeting,

    January 11-15, 1998, Washington DC, 1998

    [15] Owens D.A., Francis E.L., Leibowitz H.W., Visibility Distance with Headlights: A Functional

    Approach, SAE Technical Paper Series, Paper Number 890684, Society of Automotive Engineers,

    400 Commonwealth Drive, Warrendale, PA 15096, 1989

    [16] Andre J.T., Owens, D.A. The Twilight Envelope: An Alternative Approach Toward Describing

    Roadway Visibility at Night, In A. Gough (Ed.), Vision at Low Light Levels, Proceedings of the 4th

    International Lighting Research Symposium, pp. 171-196, May 19-21, 1998, Orlando, Florida

    [17] Adrian W., Fields of Visibility of the Nighttime Driver, Published in Proceedings of the PAL 97,

    Progress in Automobile Lighting Conference, Darmstadt University of Technology, September 23-

    24, 1997

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    30/32

    30

    [18] Zwahlen H.T., Schnell T., Target Visibility During Civil Twilight, In A. Gough (Ed.), Vision at Low

    Light Levels, Proceedings of the 4thInternational Lighting Research Symposium, pp. 171-196, May

    19-21, 1998, Orlando, Florida

    [19] Hazlett R.D., Allen M.J., The Ability to See a Pedestrian at Night : The Effects of Clothing,

    Reflectorization and Driver Intoxication, American Journal of Optometry and Archives of American

    Academy of Optometry, 45(4), 1968

    [20] Shinar, David., The Effects of Expectancy, Clothing Reflectance, and Detection Criterion on

    Nighttime Pedestrian Visibility. Human Factors. 27(3), 1985.

    [21] Chrysler, Susan T., Danielson, Suzanne M., and Kirby, Virginia M. Age Differences in Visual

    Abilities in Nighttime Driving Field Conditions. Proceedings of the Human Factors and Ergonomics

    Society. Volume 2. Fortieth Annual Meeting, 1996

    [22] Blomberg R.D., Leaf W.A., Jacobs H.H., Detection and Recognition of Pedestrians at Night,

    Transportation Research Circular Number 229, Transportation Research Board, National Academy

    of Sciences, Washington, DC, 1981

    [23] Olson P.L., Sivak M., Visibility Problems in Nighttime Driving, Authors of Chapter 15 in Automotive

    Engineering and Litigation, Garland Law Publishing, New York, 1984

    [24] Strickland J., Ward B., Allen M.J., The Effect of Low vs. High Beam Headlights and Ametropia on

    Highway Visibility at Night, American Journal of Optometry and Archives of American Academy of

    Optometry, 45(2), 1968

    [25] Austin R.L., Klassen D.J., Vanstrum R.C., Driver Perception of Pedestrian Conspicuousness, Under

    Standard Headlight Illumination, 3m Company, St. Paul, MN

    [26] Zwahlen, H.T. and Schnell, T., "Loss of Visibility Distance Due to Automobile Windshields at Night,"

    paper presented at the Transportation Research Board's 12th Biennial Symposium on Improving

    Visibility for the Night Traveler, May 23-24, 1994, National Academy of Sciences, Washington, D.C.

    Published in Transportation Research Record No. 1495, 1995, pp. 128-139.

    [27] Zwahlen Helmut T., Schnell Thomas, "The Visibility of Road Markings as a Function of Age and

    Retro-Reflectivity under Low-Beam and High-Beam Illumination at Night", paper No. 980285,

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    31/32

    31

    presented at the 77th Annual Meeting of the Transportation Research Board, Washington, DC,

    January 11-15, 1998

    [28] Zwahlen, Helmut T. and Schnell, Thomas, "Visibility of New Centerline and Edge Line Pavement

    Markings," Paper No. 971166, presented at the 76th Annual Meeting of the Transportation Research

    Board, Washington, DC, January 12-16, 1997.

    [29] Zwahlen, Helmut T. and Schnell, Thomas, "Visibility of Yellow Center Line Pavement Markings as a

    Function of Line Configuration and Line Width," paper presented at the 40th Annual Meeting of the

    Human Factors and Ergonomics Society, September 2-6, 1996, Philadelphia, Pennsyvlania.

    Published in Proceedings, 40th Annual Meeting of the Human Factors and Ergonomics Society,

    1996, pp. 919-922.

    [30] Zwahlen, Helmut T. and Schnell, Thomas, "Visibility of New Dashed Yellow and White Center

    Stripes as a Function of Material Retro-Reflectivity," Paper No. 961268, presented at the 75th

    Annual Meeting of the Transportation Research Board, January 7-11, 1996, Washington, DC.

    Published in Transportation Research Record No. 1553, 1996, pp. 74-81.

    [31] Zwahlen, Helmut T., Hagiwara, Toru, and Schnell, Thomas, "Visibility of New Yellow Center Stripes

    as a Function of Obliteration," paper presented at the 74th Annual Meeting of the Transportation

    Research Board, Paper No. 950993, January 22-26, 1995, Washington, DC. Published in

    Transportation Research Record No. 1495, 1995, pp. 77-86.

    [32] Zwahlen, Helmut T., Schnell, Thomas, and Hagiwara, Toru, "The Effects of Lateral Separation

    Between Double Center Stripe Pavement Markings on Visibility Under Nighttime Driving Conditions,"

    paper presented at the 74th Annual Meeting of the Transportation Research Board, Paper No.

    950994, January 22-26, 1995, Washington, DC. Published in Transportation Research Record No.

    1495, 1995, 87-98.

    [33] Zwahlen H.T., Schnell T., Visibility of New Pavement Markings at Night Under Low Beam

    Illumination, Transportation Research Record 1495, National Research Council, Washington,

    D.C., 1995

  • 7/21/2019 Visual Target Detection Models for Civil Twilight and Night Driving Conditions

    32/32

    32

    [34] Zwahlen, Helmut T. and Schnell, Thomas, "Visibility of New Centerline and Edge Line Pavement

    Markings," Paper No. 971166, presented at the 76th Annual Meeting of the Transportation Research

    Board, Washington, DC, January 12-16, 1997

    [35] Zwahlen, H.T., Schnell T., Legibility of Traffic Sign Text and Symbols, Paper Presentated at the

    Transportation Research Board 77th Annual Meeting, January 11-15, 1998, Washington, DC

    [36] Schnell T., Zwahlen H.T., Driver-Headlamp Dimensions, Driver Characteristics, Vehicle And

    Environmental Factors Used in Retro-Reflective Target Visibility Calculations, Paper presented at

    the Transportation Research Board 77th Annual Meeting, January 11-15, 1998, Washington DC,

    1998

    [37] Pline J.L, Editor, ITE, Institute of Traffic Engineers, Traffic Engineering Handbook, 4th Edition,

    Prentice Hall, Publishers, Englewood Cliffs, NJ, 1992