The effects of dimmable road lighting: A comparison of measured and perceived visibility

Transportation Research Part F 43 (2016) 141–156

Contents lists available at ScienceDirect

Transportation Research Part F journal homepage: www.elsevier.com/locate/trf

The effects of dimmable road lighting: A comparison of measured and perceived visibility Sanaz Bozorg Chenani a,⇑, Mikko Maksimainen b, Eino Tetri b, Iisakki Kosonen a, Tapio Luttinen a a b

Aalto University, Department of Built Environment, Espoo, Finland Aalto University, Department of Electrical Engineering and Automation, Espoo, Finland

a r t i c l e

i n f o

Article history: Received 22 October 2014 Received in revised form 6 September 2016 Accepted 4 October 2016

Keywords: Night driving Visibility Road lighting intensities Car headlights Luminance contrast

a b s t r a c t By dimming road lighting, energy can be conserved without compromising traffic safety. This paper presents a study carried out on the effect of different lighting levels from road luminaires on drivers’ visual performance on a low traffic urban road. The small uniform target was used to evaluate the visibility performance of the drivers. The results obtained from subjective graded visibility were compared with contrast and the Adrian model. Results indicated a strong correlation between subjective graded visibility and contrast (R2 = 0.94) and a positive correlation between subjective graded visibility and the Adrian model (R2 = 0.88). Target’s location in relation to road luminaires had a considerable effect on its visibility. However, visibility is not a monotonic function of road lighting level. In the absence of glare from an oncoming car, 49% (3557 lm) of road lighting intensity provided better contrast and mean visibility than 100% (7252 lm) and 71% (5179 lm) of road lighting intensities. The glare from oncoming cars reduced visibility. However, no statistically significant effect of road lighting level on visibility under glare could be found. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Detecting targets on a road is directly dependent on visibility, which in turn is related to traffic safety (Bremond, Bodard, Dumont, & Nouailles-Mayeur, 2013). Road lighting improves drivers’ visual performance at night, leading to a reduction in the number of accidents (Raynham, 2004). Wanvik (2009) estimated the effect of road lighting on accidents on Dutch roads during 1987–2006 as
49% on fatal crashes and
46% on injury crashes. In another study, Elvik (1995) reviewed 37 studies from 1948 to 1989 in different countries and discovered that road lighting is responsible for a 65% reduction in night-time fatal accidents, a 30% reduction in night-time injury accidents, and a 15% reduction in night-time property-damage-only accidents. A brief overview of 62 before and after studies in 15 countries indicated that road lighting during the night reduced the number of accidents on average by 30% (CIE, 1993). However, the positive effect of road lighting on traffic safety comes at a cost. IEA (2006) has estimated that in 2005, lighting consumed 19% of the world’s electricity, outdoor lighting amounting to about 8% of total lighting electricity consumption. The demand for road lighting during the night depends on several factors. For example, a snow-covered road requires less light than a dry road to fulfil the standard luminance requirements (luminance refers to luminous intensity per unit projected area). According to Wanvik (2009), the luminance level of a snow-covered road surface increases by a factor of 4 or 5 compared to a dry road. ⇑ Corresponding author at: Rakentajanaukio 4 A, 02150 Espoo, Finland. E-mail addresses: [email protected] (S. Bozorg Chenani), [email protected] (M. Maksimainen), [email protected] (E. Tetri), iisakki. [email protected] (I. Kosonen), [email protected] (T. Luttinen). http://dx.doi.org/10.1016/j.trf.2016.10.012 1369-8478/Ó 2016 Elsevier Ltd. All rights reserved.

142

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

Static road lighting provides a standard amount of light in all circumstances, e.g. traffic and weather conditions, while unnecessary light can be avoided when and where possible by using dimmable road lighting. Traffic safety, energy consumption and minimization of light pollution are important factors in developing new intelligent lighting designs. In Norway, for example, dimming the luminaires due to traffic, climate conditions, etc., resulted in 30% energy saving and prolonging the life expectancy of the lamps (BSREC, 2007). Hogema and Kaptein (1998) conducted a field study on the effect of dimmable road lighting on traffic behaviour and road safety. They collected data in three measurement periods: a no road lighting period, a normal road lighting (100%) period, and a dimming period (20%, 100% or 200%). The results concluded that in dry weather and low traffic volume condition, a lower lighting level (20% or 0.2 cd/m2) can be applied without having any negative road safety consequences. The headlights of oncoming cars cause glare, which in turn reduces visibility. Glare refers to the loss of retinal image contrast as a result of intraocular light scatter, or stray light (Aslam, Haider, & Murray, 2007). This causes extra light to fall on the image of the object and will lead to veiling luminance. There are four types of glare: (1) Saturation glare refers to exposure of a large part of the visual field to a long period of brightness, such as on sunny days; (2) Adaptation glare refers to a sudden and large increase in luminance of the whole visual field; (3) Disability glare refers to the scattering of light in the eye that disables the visual system to some extent; (4) Discomfort glare refers to discomfort caused by glare without necessarily impairing vision (Boyce, 2008). Many factors contribute to the effects of glare. The sensation depends on the luminance of the glare source (e.g. an oncoming car’s headlights), reflection of the road surface (e.g. from snow or ice) as well as background luminance, which all have an effect on the adaptation level of the driver, not to mention other variables on the driver’s part such as age and fatigue level. Disability glare that is caused by the headlights of an oncoming car on drivers’ vision at night is examined in an experiment conducted in this study. The effects of different lighting intensities must be studied in different scenarios on the visibility of drivers in order to determine the best level of road lighting intensity for best visibility and safer traffic conditions. Not much research is available about the combined effect of road lighting (dimmable) and car headlights. Moreover, most research has been devoted to static road lighting (no dimming), not on intelligent road lighting. Bacelar (2004) calculated the visibility level (VL) of standard targets with a reflection of 0.2 at different distances from the vehicle (a constant distance of 40 m from the vehicle for low-beam headlights and 90 m for high-beam headlights) along the axis of the road using the Adrian model (Adrian, 1989). His results indicated a variation in visibility level in (1) car headlights only, (2) road lighting only, and (3) the joint effect of headlights and road lighting conditions. VL with only headlights was constant at constant distances from the car. VL was lower at far distances using high-beam headlights than near distances using only low-beam headlights because of the lower illuminance on the target and its smaller angular size at the greater distance. VL with only road lighting was different at different locations due to light distribution. Finally, he noted that using road lighting and low-beam car headlights separately results in better visibility than when they are being used together. Ekrias, Eloholma, and Halonen (2008) studied the combined effect of car headlights and road lighting. The results indicate that the joint use of car headlights and road lighting does not improve the luminance contrast of the targets on the road. The impacts were dependent on the type of car headlights, the reflection and position of the target, the position of the car, and the road lights. The only research that considered the effect of dimming on the visibility of drivers was by Bacelar (2005). However, the research did not consider the effect of car headlights. The experiments were executed with the observer’s visibility assessment of the targets on different road lighting illuminations. The results indicated that there was a good nonlinear correlation between the decrease of luminous flux and VL. Overall, the dimming of road lights up to 50% had negligibly reduced the observer’s visibility performance. He also noted that the efficiency of the road lighting installation and the photometric characteristics of the lamp have an important effect on visibility. Installation had good longitudinal uniformity (0.7) and overall luminance uniformity (0.6). (Longitudinal uniformity refers to the ratio of the minimum to the maximum luminance along a line parallel to the length of the roadway. Overall luminance uniformity refers to the ratio of the minimum luminance at a point to the average road surface luminance over an evaluation area.) Another outcome of his study was that the position of the target had a more significant effect than dimming the lights. The goal of this study was to find the combined effect of different lighting levels and car headlights on small uniform target visibility. The effect of glare from an oncoming car on a driver’s visual performance was also estimated. This first phase of the study considers High-Pressure Sodium (HPS) lamps. The next phase, to be reported in a separate paper, considers LED lamps. 2. Visibility performance assessment Drivers’ visibility performance can be assessed in three different ways: contrast, visibility level model, and perceived visibility grade on a subjective scale. 2.1. Contrast The visibility of targets on the road depends mainly on contrast (Ekrias et al., 2008). The higher the contrast, the more visible the target becomes. The contrast of targets can be measured by both Weber’s and Michelson’s contrast, in which Michelson’s contrast is used to measure contrast in a periodic pattern (e.g. sinusoidal grating) and Weber’s contrast is used

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

143

to measure the local contrast of a single target of uniform luminance seen against a uniform background (Peli, 1990). Weber’s contrast is considered throughout this study. It can be obtained from Eq. (1):

C¼

Lt
Lb ; Lb

ð1Þ

in which C is contrast, Lt is luminance of the target, and Lb is the background luminance. Targets on the road can have a positive or negative contrast due to their polarity differences from their backgrounds: negative contrast occurs when the target appears darker than the background (negative polarity), and when the target is lighter than the background, the luminance contrast is positive (positive polarity) (Ekrias et al., 2008). According to Eq. (1), negative contrast ranges from
1 to 0 and positive contrast ranges from 0 to infinity. Consequently, contrasts of equal absolute value but with opposite signs (positive or negative) do not lead to an equal visual performance. To solve this problem, the contrast can be obtained from a redefined contrast Eq. (2), in which contrast equals the absolute value of the difference between target and background luminance divided by the larger of the two luminances (Janoff, 1992).

C¼

jLt
Lb j ; maxðLb; Lt Þ

ð2Þ

However, in the presence of glare, contrast is reduced because extra light (veiling luminance) is added to the visual field (Davoudian, Raynham, & Barrett, 2013). The amount of contrast in the presence of glare can be calculated by Eq. (3):

C¼

jLt
Lb j ; Lb þ Lv

ð3Þ

In this formula, Lv is the veiling luminance. This shows that contrast will be reduced when Lv is increased (Davoudian et al., 2013). Lv can be obtained from Eq. (4):

Lv ¼ where

k  Eglare ; hn

ð4Þ


age 4  k ¼ 9:05 1 þ 66:4  2:3
0:7 log h; 0:2 < h < 2 n¼ 2; h > 2

Here, Eglare is the illumination of glare source in lux at the eye; h (in degrees) is the angle of glare source and the fixation line valid for 1° < h < 30° from the line of sight of young drivers; and k is an age-dependent factor (CIE, 2002). 2.2. Visibility level model According to Bremond et al. (2013), the visibility level model is more relevant for traffic safety than illuminance (the amount of light striking or reaching a surface per unit area) of the road. Since increasing visibility leads to better safety, using visibility criteria for evaluating a lighting system is a logical step towards efficient lighting (Adrian, 1989). The visibility of objects on a road depends on parameters related to the object and parameters related to the observer. Object-related parameters are, e.g., contrast and size. Observers related parameters are, e.g., the observer’s visual acuity and time of observation (Bacelar, Cariou, & Hamard, 1999). Visibility level (VL) evaluated by the Adrian model is the most popular theoretical model to assess road lighting quality (Mayeur, Bremond, & Christian Bastien, 2010). VL is calculated as the ratio between the difference of luminance between the target and its background against the increment threshold of luminance. The increment threshold of luminance depends on several factors, such as the observer’s age, time of exposure, angular dimension of the object, contrast polarity (positive or negative), and adaptation luminance (Mayeur et al., 2010). A higher VL indicates higher visibility. For example, VL 8 indicates that target luminance contrast is eight times the contrast needed for object detection. Earlier studies indicate that VL should be higher than 7 in order to get sufficient visual condition (Uncu & Kayakus, 2010). VL 7 was recommended in the French guidelines for road lighting (Bremond et al., 2013). Appendices A and B describe the measurement of visibility levels by the Adrian model in the absence and presence of glare, respectively. The small uniform target can be used to weight the visibility of other targets on the road, for instance, a pedestrian. The VL of a target at one point on the road indicates the visibility of the chosen target at that point on the road. For instance, if the VL of an object at one point is zero, it does not indicate that nothing can be seen at that point, but other targets with different reflections may be seen with a higher visibility level. VL is technically meaningful, especially when it is converted to the revealing power. The term ’revealing power’ refers to the ‘‘percentage of objects, a square foot in size, placed arbitrarily on the road, which can still be seen, where the diffuse reflection factor of objects follows the statistical distribution of the clothes of pedestrians” (Narisada & Schreuder, 2004).

144

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

2.3. Psychovisual tests Psychovisual tests focus on visibility performance experienced by individuals. Participants used the subjective scale (Table 1) to grade how well they can see the target (Bremond, Dumont, Ledoux, & Mayeur, 2010; Bremond et al., 2013; Mayeur et al., 2010). 3. Materials and methods To estimate the effect of luminance level and glare on a driver’s visual performance, a set of measurements was conducted in a stationary car. A vertically mounted standard target was used: a 20 cm  20 cm uniform square with a reflection factor of 0.50, at a distance of 80 m from the stationary car (Mayeur et al., 2010). Test drivers graded the visibility of the targets (subjective graded visibility, Table 1). The results were then compared to contrast and visibility levels calculated by the Adrian model. 3.1. Measurement setups 3.1.1. Participants Five participants were selected: four males (two were 30 years old, one was 25, and one was 50) and one female (30 years old) performed a visibility grading evaluation. Each participant had a valid driver’s license. Those participants who had visual impairment wore corrective lenses to satisfy the visual requirement for driving. 3.1.2. Road A straight street section in Otaranta, Espoo, Finland, was selected for its quiet environment to perform the experiments. The road section was approximately 165 m long and 6 m wide with road markings. The road surface was wet during the experiment. In Finland, it is customary to use CIE class R2 for dry and W3 for wet road for the surface reflection properties in road lighting calculations (CIE, 1982). 3.1.3. Road lighting Road lighting consisted of five 100 W HPS lamp luminaires with a one-sided arrangement on a two-lane road. Luminaire refers to a complete lighting unit consisting of a lamp, socket and other parts that hold the lamp in place and protect it, a starter and ballast, wiring that connects the lamp to a power source, and a reflector that directs and distributes the light. Road lighting in Otaranta could be classified as class M5 according to the classification system in CIE 115:2010. The spacing between the poles was 32 m, and the height of the luminaires was 10 m. Other characteristics of the luminaires are listed in Table 2 below. The intensity of the light source could be adjusted by changing the supply voltage. Visibility was measured with 100%, 71%, and 49% light source intensities, which are equal to 230 V, 210 V, and 190 V supply voltages, respectively. Information about the effect of dimming on electrical and photometric values is listed in Table 3 below. All five poles dimmed during each measurement. The horizontal illuminance of each measurement point was measured with the lux meter. The lux meter employed was LMT (Lichtmesstechnik GMBH) Pocket Lux 2, taking into consideration just the illumination of the luminaires without the effect of car headlights. This allowed us to investigate the distribution of light in different road lighting (Fig. 1a). Vertical illuminance (lx) on the target was also measured, considering both road lighting and low-beam car headlights (in the absence of glare from the oncoming car). Fig. 1a shows that horizontal illumination is reduced as the distance to the nearest pole increases and as the road lighting intensity decreases. Targets located on the right-hand side of the roadway (right axis, points 3, 6, 9, and 12) were more illuminated (horizontally), followed by central (points 2, 5, 8, and 11) and left axes (points 1, 4, 7, and 10). Therefore, targets on the third longitudinal location from Pole 1 (points 7, 8, and 9) received the least light. These results are due to the light distribution of the luminaires as well as the geometry of the road. Whereas Fig. 1b illustrates that vertical illumination on the target increases as the target is moved closer to the second pole. It also decreased as the road lighting intensities decreased. Targets located on the left-hand side of the roadway (points 1, 4, 7, and 10) were vertically more illuminated, followed by central (points 2, 5, 8, and 11) and right axes (points 3, 6, 9, and 12). 3.1.4. Cars The two cars used for the measurements were a Volkswagen Golf and a Volkswagen Polo. The Volkswagen Polo was used as the main car and the Volkswagen Golf was used to produce glare. Table 1 Subjective visibility grading scale. Invisible

Poor

Satisfactory

Good

Very good

Excellent

0

1

2

3

4

5

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

145

Table 2 Characteristics of the luminaires.

a

Characteristics

Unit

Full level

Voltage Current Power Luminous flux CCTa Raa

V A W Lm K Ra

230 0.565 111 7252 1931 21.1

CCT is the correlated colour temperature, Ra is the colour-rendering index.

Table 3 Dimming information of lighting in Otaranta Street. Voltage/V Power/W Luminous flux/lm Luminous efficacy: lm/W

230 100% 100% 100%

210 81% 71% 88%

190 65% 49% 75%

Fig. 1. Horizontal and vertical illuminance of the measurement points in the measurement field (a) horizontal illuminance of the measurement points without car headlights, (b) vertical illuminance on the target, unit lx (lumen per square meter). The intensity of the yellow indicates the intensity of the illumination. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

146

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

32 m

Pole 1

3m

Pole 2

3

6

9

12

15

Right

2

5

8

11

14

Center

1

4

7

10

13

Left

Measurement car

80 m 70 m Glare car

Fig. 2. Measurement field.

3.1.5. Target Three targets were used for the measurement. Each was a 20 cm  20 cm uniform flat square target with a reflection factor of 0.50. The small standard uniform target represents a critical object that is difficult to perceive and dangerous for a normal-sized car (Ekrias et al., 2008; Mayeur et al., 2010). Being small in size and having no motion, colour and internal contrast, its difficult to perceive by drivers (Narisada, Karasawa, & Shirao, 2003). 3.1.6. Photometer The luminance measurements were conducted using the imaging luminance photometer TechnoTeamLMK Mobile Advanced and analysed with TechnoTeamLabSoft and Matlab. 3.2. Measurement field The measurement field was selected between two poles. The spacing between two poles was 32 m and the width of the road was 6 m (3 m for each lane of the road). One lane of this road section was divided into five lateral columns and three longitudinal rows. Each column had three arrangements (left, centre and right), and each row consisted of five arrangements. The longitudinal distance between measurement points was 8 m, and the lateral distance between measurement points was 0.75 m. The schematic diagram in Fig. 2 represents the target positions in the measurement area. For each lateral position of the targets, the measurement car remained stationary in the central axis at a distance of 80 m to the targets. In scenarios with glare from an oncoming car, the distance from the source of the glare to the measurement car was 70 m (10 m in front of the target). Although the measurement was done for all target positions (Fig. 2), the target positions under Pole 2 were not considered in this study. Target positions (under Pole 2) were higher than the previous positions (Pole 2 is 0.35 m higher than the first pole). This elevation caused the detection of targets under Pole 2 to be based on the contrast between the target’s luminance and the scenery behind the target (not the contrast between the target and the road surface, as under Pole 1). Therefore, the measurements in positions 13, 14, and 15 were not analysed. 3.3. Photometric measurements and user evaluation The measurements were performed in a stationary car with a constant distance of 80 m between the target and the car. The distance of the glare source to the measurement car was 70 m. The targets were 10 m behind the glare source in order to slightly obscure the target for drivers, thus rendering lower visibility for drivers. In the first part of the experiment, visibility of the target was evaluated as well as the effect of disability glare on the visibility. The participants sat inside the car while the targets were placed on the measurement points (each column in Fig. 2). When the targets moved to the next column, both cars were moved to keep the distance constant. Three targets with the same characteristics were placed on the measurement points at a distance of 80 m away from the measurement car (i.e., positions 1, 2 and 3 simultaneously). This allowed participants to compare the visibility of these three targets based on their visual performance and provide a more precise grade based on their comparison. All participants had to grade all targets in all positions, but the order of longitudinal locations (columns) of the targets was different for each driver. Also, dimming was not gradually represented: the order of dimming was 100%, 49%, and 71%, respectively. For each lighting level, the measurements were done in two parts. First, each driver graded the visibility of the target without the presence of glare and when he or she graded the target using the subjective graded visibility scale shown in Table 1 above. Then oncoming car lights turned on. This could affect the visibility performance of the drivers.

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

147

This procedure was repeated for each grid point under three road lighting intensities with two car headlights scenarios, one in the absence of glare (measurement car with low beam), and the other in the presence of glare from an oncoming car. In the latter case, both cars had low-beam headlights. Accordingly, each of the five participants had to grade the target visibility 72 times (12 target locations  3 road lighting levels  2 car headlight setups), resulting in a total of 360 gradings. The imaging luminance photometer was placed in the measurement car with a measuring height of 1.20 m, which is the average height of a driver’s eyes (Bacelar, 2004; Ekrias et al., 2008; Mayeur et al., 2010). The luminance of the target and its background was measured with the luminance photometer. With the information collected from measurements, the contrast and visibility level by the Adrian model were calculated. Luminance contrast in the absence of glare from the oncoming car was assessed using Eq. (2), while in the presence of glare, the luminance contrast was determined by Eq. (3), in which the theta angle was computed based on two lines. One was the line to the glare source and the other was towards the target. In experiments with the presence of glare from the oncoming car when the targets moved accordingly to the next column, both cars were moved to keep the distance constant. Since the distance of the measurement car and the glare car was the same to each lateral position of targets, the theta angle had three different values (right, centre, and left directions). Fig. 3 illustrates the theta angle for the measurements, and Table 4 displays the theta values for each direction. As Fig. 3 and Table 4 demonstrate, the theta values for different lateral positions varied, but it was constant across longitudinal positions due to constant distances. The average age of all participants (33 years old) was used in K value calculations. Also, the vertical illuminance (Eglare) that comes to the eyes decreased from 1.4 to 1.05 and to 0.89 lx by dimming the road lighting from 100% to 71% and to 49%, respectively. These values were used to calculate VL in glare scenarios (Appendix B). In VL calculations by the Adrian model, the average age of observers was 33 years (the mean age of selected drivers), and the observation time was 0.2 s. (The observation time of 0.2 s is based on results obtained from eye movement studies while driving (Schreuder, 2014). The average luminance of right and left sides of the target was measured to obtain the background luminance (Appendix A) (Adrian, 1987).) This was done for each measurement point with different pole light intensities (100%, 71%, and 49%) as well as different car headlight scenarios (glare and no-glare). To test the differences in subjective graded visibility due to different lighting levels, nonparametric Wilcoxon sign-rank and Friedman tests were conducted because the subjective graded visibility was measured on an ordinal scale. Also, a followup pairwise analysis of the Wilcoxon sign-rank test was used to compare the effects of lighting levels. A significance level of 0.05 was applied in all tests. 4. Experimental results 4.1. Contrast Fig. 4 below illustrates the contrast between the target and its background in the absence and presence of a glare source. The signs in parentheses indicate negative or positive contrast. For better visualisation, colour coding is used where blue indicates negative contrast and red indicates positive contrast. The intensity of the colours indicates the magnitude of the absolute value of the contrast. Under no-glare conditions, the target was mainly seen as positive contrast. Under full (100%) and medium (71%) lighting intensity, the contrast was negative (background lighter than the target) in positions 1, 2, 4 and 5, and in position 7 for medium intensity only. In other positions, the contrast was positive (background darker than the target). Under 49% lighting intensity, the contrast was positive in all positions. Under glare, the contrast was negative in all positions. The glare also reduced the contrast between the target and its background. 4.2. Visibility performance Table 5 lists the overall results of the average subjected graded visibilities and average visibility levels by the Adrian model for each lighting condition across all target locations. The median and standard deviation for average subjective graded visibility are given in brackets.

Fig. 3. Theta angle for positions of target on the road.

148

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156 Table 4 Theta values for different positions of the target. Theta Right Centre Left

2.46° 1.83° 1.2°

Fig. 4. Contrast between target and its background (a) in the absence of glare from an oncoming car headlight, (b) in the presence of glare. (The intensity of red indicates the reduction in contrast below 0.5.) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 5 indicates how average subjective graded visibilities change across target positions (between the poles or under the poles). Moreover, in the absence of glare from an oncoming car, visibility of target in the middle of the two poles can be considered as a critical point (position 8), in which the target was almost invisible for drivers under 71% lighting intensity. Targets located on the right, centre, and left, in that order, got better average grading in both scenarios. In the presence of glare from an oncoming car, targets located on the left received less light from road lighting and were closer to the glare source (smaller theta angle), thus they were less visible. Fig. 5 plots the effect of road lighting and glare on marginal means of grading. Estimated Marginal Means refers to the mean response of participants for each level of independent variables. The figure suggests a possible effect of lighting intensities on mean subjective graded visibility. The statistical significance of the effect of road lighting intensities and car headlights (glare, no glare) on visibility gradings was tested using the non-parametric tests. It was run three times. First, the Wilcoxon test was used to see if there was

149

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156 Table 5 Overall results of graded visibility (for mean age) and average graded visibility. Positions

Road lighting (%)

No glare

Glare

Average Visibility level

Average grading [Median, standard deviation]

Average visibility level

Average grading [Median, standard deviation]

1

100 71 49

2.44 1.83 2.01

1.6 [2, 0.55] 1.2 [1, 0.45] 1.2[1, 0.45]

0.54 0.40 0.40

0.6 [1, 0.55] 0.4 [0, 0.55] 0.2 [0, 0.45]

2

100 71 49

3.01 0.39 3.18

1.8 [2, 0.45] 0.8 [1, 0.45] 1.8 [2, 0.84]

0.91 0.10 0.30

0.6 [1, 0.55] 0.2 [0, 0.45] 0.4 [0, 0.55]

3

100 71 49

3.46 2.58 2.91

2.0 [2, 0.71] 1.6 [2, 0.89] 1.6 [2, 0.55]

0.57 1.26 0.82

0.2 [0, 0.45] 0.6 [1, 0.55] 0.4 [0, 0.55]

4

100 71 49

1.24 2.41 2.07

1.8 [2, 0.45] 1.6 [2, 0.55] 1.8 [2, 0.84]

0.65 0.48 1.16

0.8 [1, 0.45] 0.6 [0, 0.89] 0.2 [0, 0.45]

5

100 71 49

2.46 2.13 4.34

1.4 [1, 0.55] 2.2 [2, 0.45] 2.2 [2, 0.45]

1.31 0.21 0.25

0.8 [1, 0.84] 0.2 [0, 0.45] 0.4 [0, 0.55]

6

100 71 49

1.74 0.87 9.14

1.2 [1, 0.45] 1.2 [1, 0.45] 3.0 [3, 0.00]

0.52 0.40 1.25

0.4 [0, 0.55] 0.2 [0, 0.45] 1.0 [1, 0.00]

7

100 71 49

3.05 1.30 2.23

1.2 [1, 0.45] 1.2 [1, 0.45] 1.6 [2, 0.55]

0.11 0.25 0.68

0.2 [0, 0.45] 0.4 [0, 0.55] 0.2 [0, 0.45]

8

100 71 49

1.30 0.12 1.70

1.0 [1, 0.00] 0.4 [0, 0.55] 1.6 [2, 0.55]

0.86 0.19 0.13

0.8 [1, 0.45] 0.2 [0, 0.45] 0.2 [0, 0.45]

9

100 71 49

3.91 5.23 5.98

1.8 [1, 1.10] 2.2 [2, 0.84] 2.6 [3, 0.55]

1.76 1.13 1.91

0.8 [1, 0.84] 0.6 [1, 0.55] 0.4 [0, 0.55]

10

100 71 49

8.27 4.19 4.16

3.0 [3, 1.00] 2.0 [2, 0.71] 2.0 [2, 1.00]

0.70 0.64 1.40

0.8 [1, 0.84] 1.0 [1, 0.00] 0.4 [0, 0.55]

11

100 71 49

5.72 4.08 4.15

2.4 [2, 0.55] 1.8 [2, 0.84] 2.0 [2, 1.00]

1.34 1.25 0.60

1.0 [1, 0.00] 0.6 [1, 0.55] 0.8 [1, 0.45]

12

100 71 49

8.54 7.98 7.76

3.2 [3, 0.45] 3.0 [3, 0.00] 3.0 [3, 0.00]

2.87 2.52 2.41

1.0 [1, 0.00] 0.8 [1, 0.45] 0.8 [1, 0.45]

any difference between glare and no glare. Second, the Friedman test was used to test the effect of road lighting levels on subjective graded visibility in the absence of glare from an oncoming car. Third, the Friedman test was used to see if there was a statistically significant effect in subjective graded visibilities due to road lighting level in glare from an oncoming car. The effect of target positions was not tested. The signed Wilcoxon test indicates a statistically significant difference in subjective graded visibility in glare (median = 2) and no-glare conditions (median = 4), (P < 0.001). To check where the difference was, each lighting level was tested separately using the Signed Wilcoxon test. The results indicate that grading under all three road lighting intensities was statistically significant when there was glare/no-glare, in which there was a statistically significant difference in subjective graded visibility under full road lighting intensity in glare (median = 1) and no-glare conditions (median = 2), (P < 0.001). Also, there was a statistically significant difference in subjective graded visibility under medium road lighting intensity in glare (median = 1) and no-glare conditions (median = 2), (P < 0.001). And there was a statistically significant difference in subjective graded visibility under least road lighting intensity in glare (median = 0) and no-glare conditions (median = 2), (P < 0.001). The second Friedman test was conducted to evaluate differences in mean graded visibilities under full (Mean = 1.72, Std. Deviation = 0.78), medium (Mean = 1.6, Std. Deviation = 0.89) and least (Mean = 1.97, Std. Deviation = 0.80) lighting intensities in the absence of glare. The result was significant (P < 0.05). The Wilcoxon Signed-rank test was used for pairwise comparison of lighting levels. The results indicated a significant difference between mean graded visibilities under least vs. medium road lighting intensities (P < 0.05) and between full and lowest intensities (P < 0.05). There was no significant difference between full and medium road lighting intensities (P > 0.05).

150

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

Fig. 5. Relationship between car headlights and road lighting with estimated marginal means of grading.

The third Friedman test was used to test the effect of different road lighting intensities in the presence of glare. The test indicates that mean graded visibilities were not significantly different (P = 0.078). However, the slightly non-significant result may be due to the low number of participants. 4.3. Visibility level The visibility level at one point cannot describe visibility in a driving situation, but the average visibility level can be used to represent the quality of the lighting installation. Road lighting installation guidelines and standards in Europe and America vary. In Europe, road lighting design is based on illuminance and luminance. However, there are several studies in Europe that take into account required visibility at night (Zalesinska, 2012). The American standard consists of three criteria: illuminance, luminance, and small target visibility. In Small Target Visibility (STV, see Appendix C), values are given in terms of weighted average visibility level. It refers to the weighted average VL calculated by incorporating the VLs of 20 standard targets per lane placed between two luminaires; it is used as a synthesized criterion for visibility of road lighting. The size of the targets is 18 cm square with 50% reflectance. The observer is assumed to look at an angle of 1° down from the horizontal level, and each target is 273 ft (83 m) away from the observer (ANSI/IESNA, 2005). Although our measurements with slightly larger targets in 12 locations did not completely follow the requirements of the STV criteria, the weighted average VL can be used to approximately compare the results of the study with the recommended values for the target visibility level. Fig. 6 displays the Weighted Average VL for different road lighting intensities. A comparison of these values with the recommended (Table C-2) WtAvgVL values (2.2 for the local road with a medium number of pedestrian conflicts) indicates that the road lighting satisfies RP-8 regulations (ANSI/IESNA, 2005). The results of the weighted average visibility level in no-glare and glare conditions indicate similar findings as subjective graded visibilities (see Fig. 5). Fig. 6 also confirms that the effect of dimming road lighting was not monotonic. The results of statistical tests above and Fig. 6, taken together, indicate that under no-glare conditions the visibility of the target was lower for 71% and better for 49% and 100%. 4.4. Comparisons Figs. 7 and 8 displays the relationship between average subjective graded visibility with contrast and average visibility level calculated by the Adrian model. In both figures, no-glare and glare scenarios are presented by a circle and a cross, respectively, and the trend line considered both sets of data (no-glare/glare). Fig. 7 depicts the relationship between average subjective graded visibility and contrast. A linear relationship between contrast and subjective graded visibility can be seen (R2 = 0.94). Under glare, contrast and visibility are very low. But several small increases in contrast have a major effect on visibility. The experiments illustrate the importance of contrast on visibility for a driver. Fig. 8 depicts the relationship between average visibility level calculated by the Adrian model and average subjective graded visibility. Fig. 8 yields a pretty strong positive correlation between the two variables (R2 = 0.88). In other words, the visibility level calculated by the Adrian model is a very good predictor of subjective graded visibility. In the presence of glare from the oncoming car (in low visibility

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

Fig. 6. Weighted average VL (STV) for different road lighting intensities.

Fig. 7. Relationship between contrast and average subjective graded visibility by drivers (Weber’s contrast).

Fig. 8. Relationship between average visibility levels calculated by the Adrian model and average subjective graded visibility.

151

152

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

levels), there were some locations where the calculated visibility levels by the Adrian model were higher yet still had lower subjective graded visibility or lower with higher subjective graded visibility. This might be due to the sensation of glare in the eye and the position of targets on the road (the theta angle). 5. Discussion and conclusion There are two sources of artificial light for drivers during night-time: car headlights and road lighting. Drivers must have car headlights when driving at night, but road lighting may not be available all the time. These two light sources have been mostly studied separately, but there is some evidence that the combined effect of car headlights and road lighting reduces the visibility of targets on the road (Bacelar, 2004; Ekrias et al., 2008) which can, in turn, affect traffic safety. This paper reports the results of experiments on the combined effect of dimmable road lighting and car headlights on drivers’ visual performance. The effect of glare from an oncoming car that can be experienced a lot in an urban environment was also investigated. Dimming of road lighting was considered in two main scenarios: (1) in the absence of glare from the oncoming car (the measurement car had low beam headlights), and (2) in the presence of glare from the oncoming car (both cars had low beam headlights). The measurements were conducted in a stationary car. Visibility performance was assessed in three different ways: contrast, visibility level model, and subjective graded visibility. Subjective graded visibility was compared with theoretical calculations based on contrast and the Adrian model (using the luminance data of the target and its background). In this study, positive and negative contrast indicates that both car headlights and road lighting had an effect on target visibility. Graded visibility measurements (Fig. 7) also indicated that the detection of the targets depended mainly on the contrast between the target and its immediate background (R2 = 0.94). Furthermore, contrast results (Fig. 5) indicate that in the absence of glare from an oncoming car under full and medium road lighting intensities, targets were seen both in positive and negative contrast (negative contrast: positions 1, 2, 4 and 5 for 100% road lighting intensity and 1, 2, 4, 5 and 7 for 71% road lighting intensity). This means that the combined effect of car headlights and road lighting might result in zero contrast in some positions. Aleksanteri, Eloholma, and Halonen (2008) studied the effect of car headlights on target contrast in a road lighting environment, which also indicated that in negative contrast, low-beam car headlights reduced contrast, and in some cases car headlights even made the target merge into the background (zero contrast). In the current study, the effect of dimming road lighting in the proximity of car headlights on target visibility was not monotonic. Certain levels of road lighting in combination with car headlights may decrease target visibility. Thus, with car headlights the road lighting should either be bright enough or dim enough to maintain good contrast between the target and its background. The headlights of an oncoming car cause disability glare, thus reducing a driver’s visual performance. In this study, even full road lighting was not enough to ensure target visibility. The position of targets (10 m behind the glare source) made the detection of targets more challenging than if the targets were in front of the glare source. In presence on glare from an oncoming car, targets were placed 10 m behind the glare source in order to slightly obscure the target for drivers, thus rendering lower visibility for drivers. This position was selected because it considered as a critical position in glare scenarios. Participants expected to see the targets in that position. Thus even with low VLs (below and around 1 in most of the cases), they graded visibility of targets, and the subjective graded visibilities were between zero and poor. Also, the effect of age on contrast sensitivity was evident, in which the young participant (25 years old) showing the best contrast sensitivity and better subjective graded visibility. Low target visibility level values (VL < 1) are also due to the non-homogeneous illumination of the pavement (Bremond et al., 2010). As indicated in Table 5, target visibility varied according to the target position, road lighting level and the presence or absence of glare from an oncoming car. A comparison of the mean subjective graded visibility under no-glare and glare conditions confirms that glare reduces visibility. In the presence of glare, the average subjective graded visibility was between invisible and poor (M = 0.53). For instance, subjective graded visibility on the left side was lower (M = 0.48) than centre (M = 0.52) and right positions (M = 0.6). Also, it is clear that when the theta angle increases (the target is far from the glare source), the visibility level is higher than when the target had a smaller theta angle to the glare source. For instance, subjective graded visibility in the left positions (1, 4, 7, and 10) was lower than in the centre and right positions. In our study, the target visibility in glare conditions was much lower than in no-glare conditions, and the effects of different road lighting levels were not statistically significant up to 49% of road lighting. Moreover, in the presence of glare from an oncoming car, the location of the target in relation to the glare source is important (due to theta) in that when the target is far from the glare source (larger angle), the visibility is higher than when the angle is smaller. Although higher road lighting intensities were slightly better when there was glare from an oncoming car, the effect of different road lighting intensities from 100% down to 49% was not statistically significant. A prior study (Bacelar, 2004) on the combined effect of road and car headlights noted that in the absence of glare from an oncoming car, road lighting illumination is enough to satisfy the visibility of objects. The current study indicates that dimming road lighting in the presence of low-beam car headlights is feasible. Bacelar (2004) also conducted experiments on lit and unlit road to determine the effect of glare from an oncoming car. He concluded that road lighting reduces the effect of glare from the oncoming car due to the improvement in the driver’s visual adaptation. Contrary to expectations, this study did not find a statistically significant difference between different road lighting levels (100%, 71%, and 49%) under glare. The slight lack of significance (P = 0.078) may, however, be due to the relatively small sample size.

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

153

Furthermore, Adrian’s formula estimates the target visibility in laboratory conditions, which is close to the experimental conditions of these experiments. Conducting experiments on a real road provides a more sophisticated understanding of light distribution in realistic conditions. A comparison of subjective graded visibility with calculated visibility by the Adrian model (Fig. 8) shows a positive correlation (R2 = 0.88), which supports the use of the Adrian model in visibility studies on real roads. A regression function indicates that for subjective graded visibility equal to unity, VL = 0.72. These results indicate that visual performance on roads can be estimated reliably based on contrast measurements and the Adrian model. The results of these experiments can give new insight to the development of intelligent road lighting. The field measurements in this study have several challenges and limitations, including: Target: The target used in this measurement was small, uniform and stationary, but the objects in real conditions cannot always be expected to be stationary or have uniform contrast. Only one reflection factor (50%) was studied in this study. It would be useful to conduct similar measurements with different reflection factors (e.g. 20% and 70%) and a target with clothing (colour contrast). The visibility of the road environment and the road at longer distances was not studied. Lighting level of luminaries: The intensity of luminaires could be adjusted to three levels (100%, 71%, and 49%). It would be useful to have more lighting levels as well as the off-condition, in which case the influence of car headlights alone could be investigated. Car: The car was always at the same distance from the target (80 m). Different distances to the target as well as to the glare source would provide additional information. In addition, more measurements are needed to estimate the visibility when the car is moving, due to the complex task of driving. Participants: Five test drivers were used in this study. Experiments with more test drivers would provide more reliable results. However, the results obtained from these experiments were consistent. Considering the number of measurements, a larger number of participants would require more time during which lighting conditions do not change. This study has indicated that contrast measurements and the Adrian model give consistent results with the subjective visibility gradings. In many cases, subjective visibility gradings could be replaced by target luminance contrast measurements. Further research is needed for visibility levels under different weather conditions and different road lighting technologies, such as LEDs. As shown by Ekrias, Eloholma, and Halonen (2007), some weather conditions may provide a high energysaving potential. For instance, during winter in countries such as Finland, luminance levels may exceed the actual need, being even 4 or 5 times higher than under dry summer conditions. Acknowledgement The authors acknowledge the Aalto Energy Efficiency research program (project Light Energy - Efficient and Safe Traffic Environments) for financing this work. Our special thanks go to our colleague, Åsa Enberg, who helped us with traffic safety. The authors would also like to thank the anonymous reviewers for their valuable comments and suggestions. Appendix A The amount of visibility calculated by the Adrian model is the ratio between target luminance contrast against the background and the threshold contrast needed for 99.99% detection. Contrast threshold is the function of size, contrast polarity (positive or negative), background luminance, age, possible disability glare, and observation time (Mayeur et al., 2010). Visibility level under laboratory conditions is 1. When applying this figure for driving conditions, a specific threshold visibility level (known as the field factor) can be set. It is the equivalent visibility level for safe driving conditions (Bremond et al., 2013). Visibility level is evaluated by Eq. (A.1):

VL ¼

jDLactual j ; DLthershold

ðA:1Þ

where DLactual is the luminance difference between a target and its background in real conditions (Eq. (A.2)), and DLactuathershold is the luminance difference needed for minimal visibility between a target of certain angular size and its background (Eq. (A.3)).

DLactual ¼ Lt
Lb ;

ðA:2Þ

where Lt is the target luminance (cd/m2) Lb is the background luminance (cd/m2) (see Fig. A.1) If the target luminance is higher than the background, the contrast is positive. If the target luminance is lower than the background, the contrast is negative. In both cases, the minimum luminance difference for perception of the target with a certain probability level has to be evaluated by the following equation: (A.3)

pffiffiffiffi

DLthershold ¼ k

2

u pffiffiffi þ L a

aða; Lf Þ þ tg tg

F cp  AF;

ðA:3Þ

154

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

Fig. A.1. Calculation of visibility level by the Adrian model.

where k is the factor for the probability of perception (K = 2.6 for 100% probability), u is the luminous flux function (lm), L is the luminance function (cd/m2), Fcp is the contrast polarity factor, and AF is the age factor. For

Lb P 0:6

( pffiffiffiffi

u ¼ logð4:1925  L0:1556 Þ þ 0:1684L0:5867 b b pffiffiffi 0:466 L ¼ 0:05946  Lb

cd ; m2

for

Lb 6 0:00418

cd ; m2

(

pffiffiffiffi log u ¼ 0:028 þ 0:173 log Lb pffiffiffi log L ¼
0:891 þ 0:5275 log Lb þ 0:0227ðLb Þ2

for

0:00418 < Lb < 0:6

(

cd ; m2

pffiffiffiffi 2 log u ¼
0:072 þ 0:3372 log Lb þ 0:0866ðlog Lb Þ pffiffiffi log L ¼
1:256 þ 0:319logb

a: target angular
size  in minutes:

a ¼ 2  tan
1

r 60 d

Targets can be better seen in negative contrast than in positive contrast. The threshold difference between negative and positive contrast depends on background luminance and size of the target. The factor Fcp is used for this difference.

F cp ða; Lb Þ ¼ 1


ma
b ; 2:4DLpos

Contrast is 1 for positive contrast and less than 1 for negative contrast as targets are more visible. Where m comes from:

log m ¼
10
ðkðlog

Lb þ1Þ2 þ0:0245Þ

;

cd ; m2 cd k ¼ 0:075 for Lb > 0:004 2 ; m for any Lb ; b ¼ 0:6L
0:1488 b k ¼ 0:125 for Lb > 0:1

Exposure time influence: 1=2

aða; LF Þ ¼

½aðaÞ2 þ aðLb Þ2  2:1

aðaÞ ¼ 0:36
0:0972

ðlog a þ 0:52Þ

2

!

ðlog a þ 0:523Þ
2:513ðlog a þ 0:523Þ þ 2:7895 ! 2 ðlog Lb þ 6Þ aðLb Þ ¼ 0:355
0:1217 ; 2 2 ðlog Lb þ 6Þ
10:4ðlog Lb þ 6Þ þ 52:28 2

;

155

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

AF: influence of age 23 < age < 64 y

AF ¼

ðage
19Þ2 þ 0:99; 2160

64 y < age < 75 y

AF ¼

ðage
56:6Þ2 þ 1:43 116:3

Appendix B In the presence of glare, extra light scatters into the eye. To evaluate the visibility level in the presence of glare, veiling luminance is added to the background luminance to obtain the adaptation luminance:

LA ¼ Lb þ Lv ;

ðB:4Þ

The visibility level can now be calculated from Eq. (A.1). However, instead of background luminance, adaptation luminance is used. Appendix C To use STV values, the amount of pedestrian traffic in the area should be identified. This allows establishing average illuminance, average to minimum uniformity ratio, and minimum illuminance. IES-RP-8-05 defined three types of pedestrian conflict areas: low, medium, and high conflict areas. Low conflict areas are residential areas; medium conflict areas are where schools and recreational centres are located; and high conflict areas are where restaurants, shopping, and theatres are located (see Table C-1) Pedestrian conflict is assumed to be the total number of people on both sides of the street within a given section (200 m). The number also includes those people crossing the road between the hours of 18:00 and 19:00. The visibility level is used to evaluate Relative Weighted VL (RWVL) and Average RWVL (ARWVL) from Eqs. (C-1)–(C-3):

RWVL ¼ 10ð
0:1ðjVLjÞ

ðC-1Þ

Pn ARWVL ¼

j¼1 RWVLi

ðC-2Þ

n

Table C-1 Classification of pedestrian conflict: Ref: available at: http://www.saskpower.com/wp-content/uploads/ residential_streetlight_engineering_practices.pdf. Low Medium High

10 or fewer pedestrians 11–100 pedestrians Over 100 pedestrians

Table C-2 Recommended maintained values of target VL (Senadheera, Culvalci, Green, Gransberg, & Burkett, 2000). Road and pedestrian conflict area

STV criteria

Road

Pedestrian conflict area

Weighting average VL

Freeway ‘‘A” Freeway ‘‘B” Expressway

– – –

3.2 2.6 3.8

Major

High Medium Low

4.9 4.0 3.2

Collector

High Medium Low

3.8 3.2 2.7

Local

High Medium Low

2.7 2.2 1.6

156

S. Bozorg Chenani et al. / Transportation Research Part F 43 (2016) 141–156

Then the weighted average visibility level can be found by:

WtAvgVL ¼
10log10 ðARWVLÞ

ðC-3Þ

References Adrian, W. (1987). Visibility level under night-time driving conditions. Journal of Illumination Engineering Society, 16(2), 3–12. Adrian, W. (1989). Visibility of targets: Model for calculation. Journal of Lighting Research and Technology, 21(4), 181–188. Aleksanteri, E., Eloholma, M., & Halonen, L. (2008). Effects of vehicle headlights on target contrast in road lighting environments. Journal of Light & Visual Environment, 32(3), 302–314. American National Standard Practice for Roadway Lighting, ANSI/IESNA. (2005). RP-8-00. Aslam, T. M., Haider, D., & Murray, I. J. (2007). Principles of disability glare measurement: An ophthalmological perspective. Journal of Acta Ophthalmologica Scandinavica, 85(4), 354–360. Bacelar, A. (2004). The contribution of vehicle lights in urban and peripheral urban environment. Journal of Lighting Research and Technology, 36(1), 69–78. Bacelar, A. (2005). The influence of dimming in road lighting on the visibility if drivers. Journal of Light and Energy, 29(1), 44–49. Bacelar, A., Cariou, J., & Hamard, M. (1999). Calculation visibility model for road lighting installations. Journal of Lighting Research and Technology, 31(4), 177–180. Boyce, P. R. (2008). Lighting for driving: Roads, vehicles, signs and signals. New York: Taylor & Francis Group, CRC Press. Bremond, R., Bodard, V., Dumont, E., & Nouailles-Mayeur, A. (2013). Target visibility and detection distance on a driving simulator. Journal of Lighting Research and Technology, 45(1), 76–89. Bremond, R., Dumont, E., Ledoux, V., & Mayeur, A. (2010). Photometric measurements for visibility level computations. Journal of Lighting Research and Technology, 43, 1–10. Black Sea Regional Energy Centre (BSREC). (2007). Intelligent road and street lighting in Europe. Report on small scale test projects. Available at: . CIE, Commission Internationale de l’Eclairage. (1982). Calculation and measurement of luminance and illuminance in road lighting. Publication number (CIE No. 30-2 (TC-4.6)). CIE, Commission Internationale de l’Eclairage. (2010). Lighting of roads for motor and pedestrian traffic, Vienna, Technical (CIE Report No. 115, 2010). CIE, Commission Internationale de l’Eclairage (1993). Road lighting as an accident countermeasure Publication number 93(CIE No. 093-1993). Vienna: International Commission on Illumination. CIE, Commission Internationale de l’Eclairage (2002). CIE collection on glare (CIE No. CIE 146:2002 & CIE 147:2002). Vienna: International Commission on Illumination. Davoudian, N., Raynham, P., & Barrett, E. (2013). Disability glare: A study in simulated road lighting conditions. Journal of Lighting Research and Technology, 1–11. Ekrias, A., Eloholma, M., & Halonen, L. (2007). Analysis of road lighting quantity and quality in varying weather conditions. LEUKOS, The Journal of the Illuminating Engineering Society of North America, 4(2), 89–98. Ekrias, A., Eloholma, M., & Halonen, L. (2008). The contribution of vehicle headlights to visibility of targets in road lighting environments. Journal of International Review of Electrical Engineering (IREE), 3(1), 208–217. Elvik, R. (1995). Meta-analysis of evaluations of public lighting as accident countermeasure. Norway: Institute of Transport Economics. Transportation Research Record (TRR 1485). Hogema, J.H., & Kaptein, N.A. (1998). Dynamic public lighting. Phase 3: Field evaluation. Accession Number: 00809872. IEA (Light’s Labour Lost). (2006). Policies for energy efficient lighting. Available at: [Accessed 10 June 2013]. Janoff, M. (1992). The relationship between visibility level and subjective rating of visibility. Journal of the Illuminating Engineering Society, 21(2). Mayeur, A., Bremond, R., & Christian Bastien, J. M. (2010). The effect of the driving activity on target detection as a function of the visibility level: Implications for road lighting. Journal of Transportation Research, Part F: Traffic Psychology and Behaviour, 13(2), 115–128. Narisada, K., Karasawa, Y., & Shirao, K. (2003). Design parameters of road lighting and revealing power. In Proceeding of the CIE, 25th session San Diego. Vienna: CIE. Narisada, K., & Schreuder, D. (2004). Light pollution handbook. Dordrecht: Springer. Peli, E. (1990). Contrast in complex images. Journal of the Optical Society of America, 7(10), 2032–2040. Raynham, P. (2004). An examination of the fundamentals of road lighting for pedestrians and drivers. Journal of Lighting Research and Technology, 38(4), 307–316. Schreuder, D. (2014). Vision and visual perception: The conscious base of seeing. Archway publishing. Senadheera, S., Culvalci, O., Green, B., Gransberg, D., & Burkett, K. (2000). Evaluation of roadway lighting systems designed by small target visibility (STV) methods. Report number: TX-97/0-1704-8. Uncu, I. S., & Kayakus, M. (2010). Analysis of visibility level in road lighting using image processing techniques. Journal of Scientific Research and Essays, 5(18), 2779–2785. Wanvik, P. (2009). Effects of road lighting on motorways. Journal of Traffic Injury Prevention, 10(3), 279–289. Zalesinska, M. (2012). Theoretical model of the visibility level and practical means of its implementation. Journal of Design & Ecodynamics, 7(4), 381–393.