This study investigates the extent to which a windscreen affects the severity of laser eye dazzle (disability glare produced by a laser) experienced by a human observer. Windscreen scatter measurements were taken for a range of windscreens in a variety of conditions, showing that windscreen scatter is similar in magnitude to scatter from the human eye. Human subject experiments verified that obscuration angles caused by laser eye dazzle could be increased by the presence of a windscreen when comparing a dirty automobile windscreen to an eye-only condition with a 532-nm laser exposure. However, a light aircraft windscreen with lower scatter did not exhibit increased obscuration angles at 532 nm, and neither windscreen exhibited an increase at 635 nm. A theoretical analysis of laser eye dazzle, using measured windscreen scatter functions, has provided insight into the delicate interplay between scatter, transmission and the angular extent of dazzle. A model based on this analysis has been shown to be a useful tool to predict the impact of windscreens on laser eye dazzle, with the goal of informing future updates to the authors’ laser eye dazzle safety framework.
22 October 2018: Typographical corrections were made to Refs. 1, 2, and 14.
The momentary visual obscuration due to a visible wavelength laser beam being incident with the eye is known as laser eye dazzle. Commercial airline pilots are regularly subjected to such dazzle through malicious use of lasers by individuals on the ground , while ‘laser dazzlers’ are often used in military and security applications as non-lethal weapons for their ability to provide a warning signal and aid in determining intent .
A fundamental factor that contributes to the severity of laser dazzle is the extent of scatter induced by optical elements both before the light reaches the eye, and within the eye itself. In the absence of any scatter, the retinal image of the laser would be a small spot that would only affect a limited portion of the visual field. In reality, stray light from extraocular and intraocular scatter gives rise to unfocused light that is manifest as a broadened veiling glare across the scene .
Potentially, extraocular scatter can be induced by any optical elements between the laser source and the eye, including the atmosphere, windscreens and spectacle lenses. Recent studies have investigated atmospheric scatter , concluding that over short ranges (less than 400 m) it would have a negligible impact on laser eye dazzle. Research into the scatter of spectacle lenses , however, indicates that uncleaned lenses contribute significantly (> 10%) to the total scatter of the eye. This present study concerns the contribution of another extraocular scattering source – automobile or aircraft windscreens (also known as windshields, canopies or transparencies).
Allen  used photography to estimate the veiling luminance from windscreens illuminated by car headlights, in order to understand safety concerns about damaged windscreens. He concluded that old and uncleaned windscreens averaged twice as much veiling luminance as the human eye, whilst new and clean windscreens contributed less than 5% of the human eye’s veiling luminance.
Padmos  approached the issue of windscreen scatter from the perspective of the visibility of long tunnel entrances. Using a broadband light source, he measured scatter in twenty-five glass car windscreens and concluded that the scatter made a significant contribution to the overall veiling glare experienced when approaching a dark tunnel from bright surroundings.
Smith  investigated the effect of scatter from a glass aircraft windscreen and a plastic aircraft windscreen, from 1.5° to 5.5°, using an Argon ion laser (514.5 nm) and visual contrast threshold tests in human subjects. The results of these tests were converted into a windscreen scatter function that was found to be comparable to human eye scatter, with a higher scatter for the plastic windscreen due to its poorer optical quality.
Marasco and Task conducted experiments to characterize the scatter from a variety of helmet-mounted display visors and aircraft windscreens  and study the impact on visual performance . They used a broadband light source to measure haze – the percentage of transmitted light that is scattered – together with an angular profile of the scatter. They found no clear correlation between haze and angular scatter profile, while their human subject experiments confirmed that visual performance was linked more closely to the veiling luminance introduced by transparencies than the measured haze.
Blick et al.  estimated the visual effect of light scattered from a laser glare source through one canopy by combining the spatial map of the glare field (measured with a scanning radiometric imager) with a point spread function model . They found the technique to be a useful predictor of actual measurements with six human subjects at a laser irradiance of 0.5 μW·cm−2 at the eye.
More recently, Toet and Alferdinck  measured the angular dependence of laser scatter from twelve windscreens using a luminance meter and a military laser illumination device. Their results broadly agreed with the findings of Padmos, showing that used windscreens scattered more than new ones, and dirty windscreens scattered more than clean ones by a factor of around fourteen.
This present work supplements previous studies by combining scatter measurements, human subject experiments and a theoretical treatment, while exploring a much wider range of parameters in order to develop a deeper understanding of how windscreen scatter impacts laser eye dazzle. For the first time, scatter dependencies on windscreen type, surface condition, incidence angle, and laser wavelength are all explored within the same experiment. Furthermore, human visual performance in the presence of laser eye dazzle is then assessed, both with and without windscreens in place, at laser irradiances up to 600 μW·cm−2 at the eye. Finally, a theoretical analysis brings together the experimental data with a state-of-the-art dazzle prediction methodology.
The full data and analysis from this study is freely available for download to facilitate additional investigation and review beyond this article .
2.1 Scatter measurements
The angular distribution of scattered laser radiation was measured for four different windscreens, each at three angles of laser incidence, three surface conditions and two laser wavelengths. The underlying physical principles of the measurement technique are the same as those previously developed by the authors for the measurement of atmospheric scatter . This section provides a brief summary, but the reader is directed to the earlier publication for full technical details. All measurements took place during May and June 2016 at the Tri-Service Research Laboratory (TSRL) on Joint Base San Antonio – Fort Sam Houston (JBSA-FSH), Texas.
The angular distribution of windscreen scatter is defined by a scatter function that describes the fraction of light scattered, per unit solid angle, in a given direction. To measure this, a laser is propagated through a windscreen, and both the total laser irradiance transmitted by the windscreen and the amount of light scattered in a given direction are measured.
A scatter function may be determined experimentally by comparing the total laser irradiance received from all angles, to the laser irradiance received from a narrow field of view at a range of angles. First, the total transmitted laser irradiance, ITOTAL (mW∙cm–2), is measured with a Wide Acceptance Angle Detector (WAAD), located on-axis and normal to the laser path. The WAAD collects light from all angles and all directions in a hemisphere around the detector plane, gathering both on-axis and off-axis forward scattered radiation. Then, a Narrow Acceptance Angle Detector (NAAD) is used to sample the scattered light around this hemisphere and measure the contribution, I(θ) (mW∙cm–2), to ITOTAL within a fixed solid angle, Ω (sr), at a given angle, θ (°), from the laser axis. The NAAD remains on the center of the laser axis, but is rotated with a vertical axis of rotation perpendicular to the horizontal plane to replicate the way an eye would rotate to view the outside world through the windscreen. The NAAD was constructed with a 0.955° acceptance angle giving a solid angle, Ω, of 2.18 × 10−4 sr. Its sensitivity floor was around 1 nW∙cm–2.
The windscreen scatter function, fwindscreen(θ) (sr–1) is calculated as the ratio of the scattered light intensity at a given angle to the total transmitted light; the scatter fraction:
The experiment was conducted indoors using two commercially available laser sources: a green continuous wave 532-nm laser (Spectra Physics Millennia X); and a red continuous wave 635-nm laser (Laserglow Technologies LRD-0635). Collectively, green and red accounted for 93% of all laser illuminations reported to the FAA in 2017 , which demonstrates the relevance of this laser choice.
A schematic diagram of the optical arrangement is shown in Fig. 1 and a list of the primary optical components is given in Table 1. Both laser beams passed through a series of optics to provide wide, collimated beams. The 532-nm laser beam was first expanded 12 × to provide a 28 mm beam that more closely matched the 27 mm output from the fiber collimator of the 635-nm laser. A second beam expander then provided an additional 12 × expansion of both laser sources, after which the beam was directed onto a 500 mm diameter spherical collimating mirror with a focal length, f, of 4000 mm. The final collimated output beam, which was 100 mm in diameter, was reflected off a plano turning mirror before it was incident on the front surface of the windscreen.
The Design Eye Reference Point (DERP – the optimum position for the pilot/driver’s eyes in a given vehicle) and the rake angle for each windscreen were determined from platform measurements and analysis of photographs. The NAAD was positioned at the DERP appropriate for each windscreen, mounted at the center of rotation of a rotation stage (Velmex 4800) that allowed for angular movements as small as 1/1280th of a degree. The rotating NAAD assembly is illustrated in Fig. 2.
The maximum measurement angle defined by the diameter of the incident beam on the windscreen was limited to 4°, but an experimental adjustment allowed measurement out to 40°. The turning mirror was mounted onto a translation stage, which allowed the beam to be moved laterally across the windscreen (Fig. 3). The lateral motion of the beam was then synchronized with the rotation of the NAAD, so the detector’s 1° field of view was steered onto the point where the center of the laser beam was incident upon the windscreen (Fig. 4). This allowed a maximum measurement angle of 40° (limited now by the geometry of the windscreen), with measurable scattered light well above the noise floor of the NAAD. A calibrated laser power detector (Newport 2936-R Benchtop Optical Power Meter) was used to measure the laser irradiance transmitted through the windscreen, and this served as the WAAD.
Critical to the scatter measurement system was the inclusion of a fixed obscuration disc in front of the NAAD, designed to prevent the direct beam from entering the lens of the NAAD, so that only scattered light was sampled. The size of the disc was slightly larger than the NAAD aperture at approximately 25 mm, which facilitated alignment by casting a shadow across the NAAD aperture. The obscuration disc was mounted on an alignment table, 4 m away from the NAAD (see Fig. 3), to ensure that it did not encroach upon the NAAD’s field of view and block any scattered light that was part of the desired measurement. Measurements began at an angle of 1° to ensure that the obscuration disc was not within the detector’s field of view, and the Arago spot (the bright spot at the center of the disc’s shadow caused by Fresnel diffraction) had no measurable impact on the detector.
Also mounted on the alignment table was a HeNe laser, which allowed accurate alignment of the NAAD with the laser beam. With a plano-mirror fixed onto the front surface of the NAAD, the HeNe laser was aligned to create two retro-reflections; one from the front surface of the NAAD and the other from the front surface of the windscreen. Windscreens were mounted on a custom-made aluminum frame that allowed for easy rake angle adjustment, and the position of the windscreen and NAAD were adjusted so that both retro-reflections were co-aligned with the forward laser beam path. This position was also used as a ‘zero’ reference calibration point for a clinometer that was coupled to the center of rotation of the windscreen mount and used to set the windscreen to the desired rake angle.
Four windscreens were used in this study (see Fig. 5): an automobile windscreen from a 2014 Chevrolet Malibu (subsequently referred to as ‘Laminated Glass’); a clear windscreen (‘Clear Acrylic’) and tinted windscreen (‘Tinted Acrylic’) from a Cessna 172 aircraft; and a highly curved ‘bubble’ type windscreen from a Zenith Zodiac 601 kit aircraft (‘Bubble Acrylic’).
Measurements were made for three rake angles: ‘normal’, ‘as-flown/driven’ and ‘extreme’. For the ‘normal’ (N) incidence angle, the laser beam was normal to the windscreen surface, representing a laser impact from above the platform such as a car being illuminated from an overhead bridge. ‘As-flown/driven’ (F) represents an impact from the same height as the windscreen such as a ground-based illumination of a car, while the ‘extreme’ (E) angle represents a laser engagement from below, such as a ground-based laser targeting an aircraft. Rake angles are reported as the angle from the vertical, where a plano windscreen would have a 0° rake angle when stood vertically and a 90° rake angle when lying horizontally.
Each windscreen was measured in three conditions of physical cleanliness; ‘as-is’, ‘clean’, and ‘dirty’. The ‘as-is’ condition was with the windscreen in its initial state as received at the laboratory. The ‘clean’ condition was accomplished by applying a glass cleaning solution to the windscreen and thoroughly wiping any particles from both the inner and outer surfaces with a microfiber cloth. The ‘dirty’ condition involved the application of particles to the outside surface to simulate the accumulation of dirt on a windscreen. For this, two light coatings of a suspension of white developing particles in a fast drying solvent (Magnaflux Spotcheck SKD-S2) were sprayed onto the front surface of the windscreen. The optical scatter induced by these particles resembled what one might observe through a used, dirty windscreen, although a comparison to the particle size and concentration of ‘real’ dirt was not performed.
Scatter measurements were made for each combination of windscreen, wavelength, rake angle and condition as detailed in Table 2. NAAD rotation angles ranged from 0° to 40° at 0.25° increments, with 10,000 irradiance samples being averaged for each measurement (1 s of data gathering at 10 kHz).
2.2 Human observer experiments
Twelve human observers were exposed to laser eye dazzle, with test conditions including two different windscreens plus a no-windscreen baseline condition, across two different laser wavelengths. The basic methodology of the experiment is the same as the authors’ previously reported human subject dazzle experiment . A short summary is provided in this section, but the reader is referred to the earlier publication for more complete details. All human observer experiments took place during November and December 2016 at the TSRL on JBSA-FSH, Texas.
The experiment was designed to determine the angular extent of laser eye dazzle using a simple letter orientation task. During exposure to laser light, human subjects were asked to determine the orientation of ‘E’ characters (0.4° in size with 60% contrast) presented at different angular eccentricities, θ, from the laser axis. These angular eccentricities were spaced in 0.2 log steps, comprising angles of 31.6°, 20.0°, 12.6°, 7.9°, 5.0°, 3.2°, 2.0°, 1.3°, and 0.8° (dictated by the smallest achievable viewing angle).
By moving from large to small eccentricities, a threshold angle can be found at which the observer can no longer recognize the orientation of the ‘E’ due to obscuration from the laser dazzle. With a widest presentation eccentricity of θmax (°), a total number of correct responses across all eccentricities of Xsum, and a maximum possible number of correct responses for each eccentricity of Xmax, a threshold eccentricity value, θth (°), can be calculated according to the following equation:16]. The eccentricity was decreased until the subject felt the laser dazzle was too bright to guess. The resulting threshold eccentricity corresponded to the smallest angle at which the observer would be expected to correctly recognize the orientation of the presented ‘E’ character for the given dazzle condition.
‘E’ characters were presented to the subjects on a tablet computer (Samsung Galaxy Tab 4) attached to a movable robot on a curved rail (see Fig. 6). This is the same implementation as the authors’ earlier outdoor experiment  although the present experiment was conducted indoors at an ambient and tablet luminance of 10 cd∙m−2. Subjects were situated with their head supported by a chin rest with monocular (right eye) viewing of negative contrast letters on the tablet. The ‘E’ character was presented first at the widest eccentricity four separate times with a random orientation, with three or four correct responses resulting in the robot moving the tablet to the next eccentricity for four more random presentations. With two incorrect responses for a given eccentricity, no lower eccentricities were attempted and a new scenario was then initiated.
The Laminated Glass and Clear Acrylic windscreens were used for this experiment in ‘as-flown/driven’ angles and in a ‘day of test’ condition, together with an ‘eye-only’ condition involving no windscreen. These two windscreen types were selected as they are the most commonly-used of the four types assessed in the scatter measurement experiments. Both windscreens were left in the ‘dirty’ condition after the scatter measurements for a period of approximately 6-months, during which time they accumulated additional dust in a busy indoor laboratory environment. Their condition was therefore designated as ‘day of test’ to recognize that different scatter would be expected from the previously measured conditions. Prior to each session of human subject experiments, scatter measurements were taken according to the procedure detailed in section 2.1.2 so that this could be accurately monitored. These ‘day of test’ scatter measurements are provided in the accompanying data set .
For each of the scenarios in Table 3, threshold eccentricities were determined for laser irradiances of 1.9, 6, 19, 60, 190 and 600 µW∙cm−2 delivered to the subject’s eye, using the same green (532 nm) and red (635 nm) laser wavelengths from the earlier measurements (with the exception that the red laser did not have enough power output to achieve 600 µW∙cm−2 at the subject’s eye). The laser beam was expanded further to give a large diameter beam that could be simultaneously incident upon the eye and the portion of the windscreen at the widest 31.6° eccentricity. For the windscreen cases, the irradiance at the front of the windscreen was higher in order to deliver the stated irradiances at the eye after undergoing transmission losses through the windscreens (approximately 80% for the Clear Acrylic at 532 nm and 635 nm, 70% for the Laminated Glass at 532 nm, and 50% for the Laminated Glass at 635 nm).
Twelve human volunteer subjects participated in the experiment, with an average age of 42.2 ± 17.3 years and an average eye (iris) pigmentation of 0.58 ± 0.45 (where eye pigmentation is quantified as 0 for very dark, 0.5 for dark, 1.0 for light and 1.2 for very light eyes ). All subjects had visual acuity correctable to 20/20 in each eye (distance vision correction was permitted), normal color vision (Pseudoisochromatic Plates), a normal Amsler grid, and no signs of active ophthalmic disease to include abnormal lens opacity (i.e. cataracts). These criteria were verified via a comprehensive eye exam by a licensed optometrist or ophthalmologist conducted prior to the start of data collection. The study was conducted under the tenets of the Declaration of Helsinki and informed consent was obtained from all subjects before participating in the study. All procedures were conducted in accordance with the policies of the United States Air Force .
3. Results and discussion
3.1 Scatter measurements
3.1.1 Example scatter data
Figure 7 shows a typical scatter measurement from these experiments, in this case from the Clear Acrylic windscreen in clean condition with normal angle of incidence for a 532-nm laser. It is the average of seven separate measurements, with error bars representing the average standard deviation. The decrease in scatter with angle follows a similar trend to those observed for light scatter in the eye  and in the atmosphere . Localized fluctuations from the general trend can be attributed to areas of the windscreen that have imperfections such as scratches that give regions of high scatter. Most subsequent scatter graphs will be presented with the same scales for ease of comparison, and average standard deviation error bars will be left off when comparing multiple scatter profiles for ease of viewing.
Figure 8(a) shows the highest and lowest windscreen scatter results as measured by these experiments. The highest scatter was measured for the Clear Acrylic windscreen in dirty condition at a normal angle of incidence for a 532-nm laser. The lowest scatter was for the Laminated Glass windscreen in clean condition at normal angle for a 532-nm laser. Figure 8(b) puts these into context on an expanded scale by comparison to human eye scatter and atmospheric scatter. Human eye scatter is derived from the authors’ earlier work , assuming a 40 year old observer with dark eyes, with the high ambient luminance eye scatter case based upon daytime ambient light levels (1,000 cd·m−2) and the low ambient luminance eye scatter case assuming nighttime ambient light levels (0.1 cd·m−2). The atmospheric scatter is based upon the authors’ earlier work  which involved measurements over a 380 m laser propagation path in good to moderate air quality.
It can be seen that windscreen scatter, even at the lowest measured level, is considerably higher than the atmospheric scatter – around three orders of magnitude higher at most angles. When compared with eye scatter, the highest measured level of windscreen scatter is near the high ambient luminance eye scatter. The lowest windscreen scatter is lower than even the low ambient luminance eye scatter, although it reaches about 10% of the low ambient luminance eye scatter value at 7°, 20% at 15°, and approaches 50% around 35°. Since scatter in the eye is the primary source of laser glare, these comparisons suggest that windscreen scatter could have an impact on the overall laser dazzle severity, as will be explored later in this paper.
Figure 9 compares the highest windscreen scatter measured in this work to that measured by Padmos  (also verified by Toet and Alferdinck ) and Smith . Both Padmos and Smith measured unclean windscreens that had undergone real use, and therefore the most relevant comparison to this work is to a high scatter case as none of the windscreens under test had experienced considerable real-world wear. It is notable how similar these measurements are, despite the very different experimental techniques: Padmos used direct luminance measurements with a broadband light source; Smith derived his scatter functions from human subject laser exposure experiments; and this work measured laser irradiance directly within the beam axis. Fitting equations for the highest and lowest scatter cases from the present measurements are listed in Table 4 together with those reported by Smith and Padmos. Similar fitting equations have historically been applied to eye scatter functions, with parameters for the classic Stiles-Holladay disability glare formula also given in Table 4 .
The remainder of this section will investigate comparisons within the windscreen scatter data to reveal trends and dependencies.
3.1.2 Windscreen type
Of the four windscreens measured, the Bubble Acrylic windscreen generally gave the highest scatter values and the Clear Acrylic and Laminated Glass windscreens generally gave the lowest. Figure 10(a) compares the scatter from all four windscreen types in clean condition, with the 532-nm laser at the as-flown/driven angle. The Bubble Acrylic windscreen exhibited the highest scatter at all angles, while the Tinted Acrylic begins with the second highest scatter at narrow angles (up to 2°) before falling within the same region as the Clear Acrylic and Laminated Glass windscreens. Figure 10(b) shows the same comparison but with artificial dirt on all windscreens, which appears to even out the scatter performance although the Laminated Glass consistently exhibits the lowest scatter. This agrees with the findings of Smith  who reported lower scatter from glass windscreens, with the Laminated Glass being the only glass windscreen on test here.
3.1.3 Surface condition
All windscreens exhibited higher scatter when the artificial dirt was introduced compared to their clean condition, as would be expected. Figure 11 gives an example of this trend for the Clear Acrylic windscreen, showing a 4 to 24-fold increase in scatter for the artificial dirt case, depending upon the angle considered. This provides a reasonable match to the 14-fold increase for dirty versus clean windscreens reported by Toet and Alferdinck . A comparison of all windscreens in their original ‘as-is’ condition is not relevant as each had different levels of natural dirt/dust build-up based upon their history.
3.1.4 Laser incidence angle
The angle of the windscreen with respect to the laser was found to have very little impact on the measured scatter. Figure 12(a) illustrates that incidence angle had no impact on scatter for the Clear Acrylic windscreen in clean condition, while Fig. 12(b) shows only a very slight increase in scatter with increasing angle of incidence for the Laminated Glass windscreen in dirty condition. No large disparity in scatter values was witnessed in any of the test cases, with differences typically within their uncertainties, leading to the conclusion that the angle of incidence has a negligible impact on the amount of windscreen scatter. This is in agreement with the findings of Padmos  and Toet and Alferdinck .
While scatter was not affected by incidence angle, it should be noted that the transmission of a windscreen will be affected due to the increased reflections with increased rake angles. Walsh  demonstrated how rake angle reduced transmission for car windscreens, showing a drop from around 90% at 0° to 80% at 60° and around 65% at 70°. A reduced transmission reduces the amount of light available for scattering and therefore will reduce the severity of laser eye dazzle.
3.1.5 Laser wavelength
For most cases the green (532 nm) laser light was found to exhibit slightly higher scatter than red (635 nm) laser light at small angles, and slightly lower scatter at high angles. Figure 13 shows the case for the Clear Acrylic and Laminated Glass windscreens in clean condition in the as-flown/driven orientation, where the cross-over point was around 5° in both cases. However, these differences in scatter between the two wavelengths were typically within their standard deviations and can therefore be considered to be negligible.
3.2 Human observer experiments
3.2.1 Threshold angle vs. laser irradiance at the eye
Figure 14 summarizes the results of the human subject experiments by plotting the threshold angle as a function of the laser irradiance at the eye. Threshold angles were derived by log-averaging threshold eccentricities across all subjects, with 95% confidence intervals derived using the t-distribution. It can be seen that the introduction of a windscreen increases the angle to identify the ‘E’ character orientation compared to the no windscreen condition. The red laser was much less effective than the green laser, with no data presented for the red 1.9 μW·cm−2 irradiance level as there was not enough dazzle to obscure the ‘E’ character for some of the subjects.
To test for significant differences, a 2 X 3 X 4 repeated measures ANOVA was performed using Greenhouse-Geisser epsilon-adjusted degrees of freedom. The factors were laser (532 and 635 nm), windscreen (no windscreen, Clear Acrylic, and Laminated Glass), and laser irradiance at the eye (6, 19, 60, and 190 µW∙cm−2 were common across the two lasers). Threshold angles were larger for the 532 nm laser than for the 635 nm laser; F(1, 11) = 217.31; p < 0.001. There was a main effect of windscreen; F(1.95, 21.44) = 14.937; p < 0.001. The Laminated Glass windscreen had a large impact relative to no windscreen (F(1, 11) = 25.763; p < 0.001), but the impact of the Clear Acrylic windscreen only approached statistical significance (F(1, 11) = 4.019; p = 0.070). The threshold angle increased with laser irradiance (F(1.96, 21.52) = 376.281; p < 0.001) with a linear relationship (F(1, 11) = 1225.364; p < 0.001). Differences between windscreens were larger with the 532 nm laser than with the 635 nm laser; F(1.96, 21.52) = 18.367; p < 0.001. The impact of the Laminated Glass windscreen relative to no windscreen was larger with the 532 nm laser than with the 635 nm laser (F(1, 11) = 28.051; p< 0.001), but there was no interaction between laser and the impact of the Clear Acrylic windscreen relative to no windscreen (F < 1). No other interaction was significant.
A separate repeated-measures ANOVA was computed for each laser (532 nm across six laser irradiances, and 635 nm across four laser irradiances). For the 532 nm laser, there was a main effect of windscreen; F(1.90, 20.94) = 54.629; p < 0.001. The Laminated Glass windscreen had a large impact relative to no windscreen (F(1, 11) = 83.913; p < 0.001) and the Clear Acrylic windscreen also had a large impact relative to no windscreen (F(1, 11) = 8.961; p = 0.012). Threshold angle increased with laser irradiance (F(2.11, 23.20) = 438.733; p < 0.001) with a linear relationship (F(1, 11) = 940.958; p < 0.001). Figure 14 shows that, with the 532 nm laser, the slopes are the same across the three windscreens and that the intercept for the Laminated Glass and Clear Acrylic windscreens are higher than for the no windscreen. For the 635 nm laser, there was no main effect of windscreen; p = 0.264. Threshold angle increased with laser irradiance (F(2.37, 26.07) = 344.126; p < 0.001) with a linear relationship (F(1, 11) = 1507.753; p < 0.001). The windscreen by laser irradiance interaction approached statistical significance; F(3.22, 35.41) = 2.289; p = 0.091. The impact of the Laminated Glass windscreen compared to no windscreen increased with laser irradiance; F(1, 11) = 5.281; p = 0.042. Figure 14 shows that, with the 635 nm laser, there were no differences between the two lower irradiances, but that the Laminated Glass windscreen produced larger threshold angles at the higher irradiances.
3.2.2 Threshold angle vs. laser irradiance at the front of the windscreen
To analyze the impact of introducing a windscreen to a given laser engagement scenario, the laser irradiances need to be equated at the front of the windscreen rather than at the eye. This is because the transmittance of the windscreen reduces the laser irradiance at the eye relative to a no windscreen condition. As detailed earlier, transmission was approximately 80% for the Clear Acrylic at 532 nm and 635 nm, 70% for the Laminated Glass at 532 nm, and 50% for the Laminated Glass at 635 nm.
Figure 15 and Table 5 summarize the results of the human subject experiments, together with theoretical model outputs to be detailed in section 3.3, for the case of equal irradiance at the front of the windscreen rather than at the eye. The tabular human subject data were derived at three laser irradiance levels (5, 50 and 500 µW∙cm−2) by fitting a power curve to the experimental data to interpolate intermediate points. In each case, the stated laser irradiance is the value at the front of the windscreen – in the eye-only case this is the same as the irradiance reaching the eye, whereas with a windscreen in place the irradiance reaching the eye is reduced by the windscreen’s transmission.
Using the 532-nm laser, threshold angles with the Clear Acrylic windscreen in place were nearly identical to threshold angles with no windscreen. In both cases the threshold angle was 3° for 5 µW∙cm−2, increasing to 8° at 50 µW∙cm−2 and increasing to 22° for the eye only case, and 23° with the Clear Acrylic windscreen at 500 µW∙cm−2. While the Clear Acrylic windscreen is adding an additional source of scatter, its transmission of 80% is actually reducing the irradiance to the eye, which therefore reduces the effect of the eye’s own scatter compared to the case without a windscreen. It appears that the scatter from the Clear Acrylic windscreen compensates for this 20% loss at the eye, but not enough to make the dazzle worse than the eye-only case for an irradiance of 50 µW∙cm−2. At higher irradiances, there is a small effect and the angle is larger with the windscreen than without the windscreen.
With the same 532-nm laser, the presence of the Laminated Glass windscreen increased threshold angles moderately at low irradiances but to a greater amount at higher irradiance levels when compared to the no windscreen and Clear Acrylic conditions. Its threshold angles were 4° for 5 µW∙cm−2, increasing to 11° at 50 µW∙cm−2 and 28° at 500 µW∙cm−2. This implies that the Laminated Glass windscreen’s scatter is sufficient to overcome the 70% transmission loss, and therefore results in an overall increase in the laser eye dazzle experienced compared with the no-windscreen case.
To visualize what this means in terms of the qualitative impact of laser eye dazzle on an observer, Fig. 16 shows a simulation of the dazzle effect increasing from 8° to 11° at 50 µW∙cm−2, and from 22° to 28° at 500 µW∙cm−2 (i.e. comparing the no windscreen case to the Laminated Glass case). Note that these threshold angles are measured from the center of the laser source, so they represent the radius of a circular dazzle field. These visualizations are based upon an extended version of a previously reported technique . They use the scatter function generated by previous human subject experiments  and incorporate additional visual effects for added realism, such as a representation of the ciliary corona .
For the 635-nm laser, all three cases gave the same threshold angles after fitting a power curve to their data, with values of 1° for 5 µW∙cm−2, 3° at 50 µW∙cm−2 and 8° at 500 µW∙cm−2. This implies that the transmission of each windscreen was cancelling out the additional scatter that each windscreen introduced at these angles.
3.3 Theoretical analysis
The impact of windscreen scatter on laser eye dazzle can be understood via the veiling luminance added to the scene due to the windscreen and the eye itself. For a given scatter function, f(θ) (sr−1), the equivalent veiling luminance, Lv (cd·m−2), caused by laser light can be expressed as :
For the case of an eye being exposed to a laser with no windscreen in place, the total veiling luminance becomes:15,17]:
With a windscreen in place, the irradiance at the eye is reduced by T, the optical transmission of the windscreen at the laser wavelength, and additional veiling luminance is introduced by the windscreen to give a total veiling luminance of:Table 4.
Figure 17 illustrates the impact of introducing a windscreen by comparing feye(θ) to (fwindscreen(θ) + T∙feye(θ)) as extracted from a comparison between Eqs. (5) and (8). The figure presents each of the test conditions used for the human subject experiments, with values for fwindscreen(θ) derived experimentally from measurements of the Clear Acrylic and Laminated Glass windscreens prior to each human subject experimental session. As in section 3.1.1, these were then averaged across all measurements, totaling four for the Clear Acrylic windscreen with each of the green and red lasers, and five for the Laminated Glass windscreen with each laser. Table 6 lists the resulting fitting equations for each windscreen scatter function, together with their transmission values. Inputs for feye(θ) to Eq. (6) were the ambient luminance for the experiments, (Lb = 10 cd·m−2), and the average age (A = 42.2) and eye pigmentation (p = 0.58) of the human subjects. The subsequent analysis is specific to these parameters, but would be different for alternative parameters (e.g. for subjects who are significantly younger or older than the average age used here).
From Fig. 17 it can be seen that, at 532 nm, the scatter of the eye-only case is very similar to the Laminated Glass windscreen out to around 5°, and to the Clear Acrylic windscreen to out around 10°. The more significant divergence comes beyond 10° where the scatter functions imply a moderate increase in scatter with the windscreens, with a slightly higher scatter predicted from the Laminated Glass windscreen than the Clear Acrylic. At 635 nm, both windscreens are similar to the eye-only case until beyond 5° when a more significant increase in scatter can be seen, with the Clear Acrylic windscreen predicted to give greater scatter than the Laminated Glass in this case.
These scatter curves can be used to predict laser eye dazzle obscuration angles by solving Eqs. (5) and (8) for laser irradiance, U, using the eye and windscreen scatter functions as previously described, together with the appropriate windscreen transmission and photopic luminous efficiency values (V(λ) = 0.88 at 532 nm and 0.24 at 635 nm). The equivalent veiling luminance, Lv, was calculated using the following relationship as derived in the authors’ earlier work ,15] with an observer age of 42.2, which led to an Lv value of 50 cd·m−2.
Figure 15(b) and Table 5 show the outputs from the model as predicted threshold angles for given laser irradiance values at the front of the windscreen. At 532 nm, the predicted threshold angles are within 1° of the experimental values (Fig. 15(a) and Table 5) for exposures up to 50 µW∙cm−2 and resulting obscuration angles up to 10°. The model is similarly accurate for 635 nm exposures at these levels. At higher laser irradiances the model does not perform as well, with 500 µW∙cm−2 exposures predicted to exceed the 40° limit of the model’s applicability for both green and red windscreen exposures. Some of this discrepancy may be attributable to the beam profile in the human subject experiments, which resulted in the laser irradiance being lower at wider angles than on-axis. The model assumes the same laser intensity across the entire beam, (as was the case with the scatter measurements), which would therefore over-estimate the additional veiling luminance at wider angles. However, the general trends of the graphs in Fig. 15(b) clearly follow those of the experiment. Notably, they show that the Laminated Glass has the highest threshold angles in the 532 nm case, which was true of the experimental findings.
Referring back to the scatter functions in Fig. 17, the model predictions do follow the anticipated trends. The Laminated Glass 532 nm exposures give higher threshold angles than the Clear Acrylic beyond 10° as the scatter graphs implied. The 635 nm windscreen exposures are also very similar at low obscuration angles as the scatter graphs of Fig. 17 implied. The scatter analysis showed that threshold angle differences at 635 nm would become more pronounced at larger angles, but the experiment only caused obscuration angles out to 8° which is why this difference was never seen. Again, it should be stressed that these observations are only valid for the average test conditions being compared, and would be different for alternative conditions such as older or younger subjects.
This study has found that windscreens do have an impact on laser eye dazzle, but the magnitude of this impact depends upon a delicate interplay between scatter, transmission and the angular extent of dazzle.
An experiment to measure windscreen scatter has been shown to be robust through comparison to the literature, while revealing new insights to scatter dependencies. The Clear Acrylic and Laminated Glass windscreens generally had lower scatter than the Tinted Acrylic and Bubble Acrylic windscreens for most conditions, while all windscreens had higher scatter in a ‘dirty’ state versus a ‘clean’ condition. The laser incidence angle was found to have little impact on the magnitude of scatter. In terms of laser wavelength, green (532 nm) laser light generally gave higher scatter than red (635 nm) at small angles (< ~5°) and lower scatter at larger angles. The highest scatter measured by this study was shown to be comparable to the scatter from the human eye. However, the windscreens in this work were all unused, and even higher scatter would be anticipated from windscreens that have experienced real-world conditions.
Human observer experiments showed some increases in the severity of laser eye dazzle due to the presence of a windscreen. Comparing eye-only to Laminated Glass at 532 nm, this manifested as an increase in threshold angle from 3° to 4° at 5 µW∙cm−2 in front of the windscreen, 8° to 11° at 50 µW∙cm−2, and 22° to 28° at 500 µW∙cm−2. However, there was no noticeable impact on laser eye dazzle with the introduction of the Clear Acrylic windscreen for the 532 nm exposures. At 635 nm there was no change in threshold angles for exposures up to 500 µW∙cm−2 comparing the eye-only and windscreen cases, with maximum obscuration angles of 8°.
A modelling approach, using experimentally measured windscreen scatter functions, has provided greater insight to the mechanisms through which windscreens impact laser eye dazzle. The presence of a windscreen during a laser eye dazzle event gives additional scatter, but it also introduces transmission losses that reduce the irradiance delivered to the eye. It is this balance of additional scatter but reduced irradiance that affects whether laser eye dazzle is made more or less severe by the introduction of a windscreen.
Modelling of the human subject experiment has shown that the windscreen scatter and transmission combined to give an overall increase in scatter versus the eye-only condition. It is this increase in total scatter that leads to an increased severity of laser eye dazzle. However, the effects became more prominent with increasing angle. For example, with 635-nm laser exposures it was shown that the scatter differences are most pronounced beyond 5-10° which was never reached in the experiment. This explains why the experiment did not find an increase in threshold angle at 635 nm with the windscreens compared to the eye-only case.
Overall, the model was shown to be a useful tool to predict the impact of windscreens on laser eye dazzle for the given test conditions, although its accuracy was best at laser irradiances below 50 µW∙cm−2 at the front of the windscreen. The model predicted very similar threshold angles to the experiment for 532 nm and 635 nm exposures of 5 µW∙cm−2 and 50 µW∙cm−2, but the discrepancies increased at exposures of 500 µW∙cm−2. The model also predicted the experiment’s findings that the Laminated Glass windscreen had the highest threshold angles in the 532 nm case and the lowest in the 635 nm case compared with the Clear Acrylic windscreen.
For the reader’s own application, this study has provided a robust methodology for characterizing windscreen scatter and predicting the resulting impact on laser eye dazzle, while giving guidance on the anticipated trends. Furthermore, Table 4 and Table 6 provide typical windscreen scatter equations that can be used in such calculations, while complete scatter data has also been made available . The following section details a roadmap towards providing more universal advice on windscreen scatter and its impact on laser eye dazzle.
5. Additional considerations and future work
The impact of windscreen scatter will vary depending upon the viewing conditions and the person who is experiencing laser eye dazzle. Human eye scatter depends upon the ambient light level (higher scatter at higher ambient light level), the observer’s age (higher scatter with older age) and eye pigmentation (higher scatter with lighter eyes). Windscreen scatter has none of these dependencies, and so the significance of its contribution will vary depending upon the baseline eye scatter for the given situation. For conditions with low eye scatter (i.e. low ambient light level, young observers, or dark eyed observers) this could cause the impact of windscreen scatter to be more severe.
The model used in this study assumes that contrast reduction due to the windscreen’s additional veiling luminance is the only additional cause of visual loss, but as Marasco and Task highlighted  there are several other potential contributors. These include the illuminated windscreen attracting the eye’s focus, and the disruptive nature of a pattern being superimposed on the target being viewed. The human subject experiments automatically include the impact of these effects, but they would need to be added explicitly to the model to have any impact there.
This work has only considered the impact of windscreens on laser eye dazzle where the laser is incident upon the eye, but windscreens also have an impact when this is not the case. For example, a laser tracking an aircraft only needs to hit the windscreen to cause a visual issue for the aircrew, even if they do not view the laser directly. That is because the windscreen scatters the laser light and masks the aircrew’s vision through the region of the windscreen being illuminated, as well as potentially adjusting their level of dark adaptation.
An obvious extension to this study is the use of a wider range of windscreens in a more diverse range of conditions. It would be particularly interesting to investigate how well-used (i.e. scratched and worn) windscreens compare in terms of their scatter and transmission characteristics. Experimental measurements could be taken and used as inputs to the model, thus removing the need for further human subject experiments. The creation of reproducible scattering sources would also be beneficial to such future research; allowing cross-comparison of studies to a set of well-characterized baseline conditions.
Should further human subject experiments be possible, however, there are a number of further advancements that could help to develop a deeper understanding. Conducting the experiment at higher and lower ambient light levels would allow additional model validation. Lower ambient light levels would be of particular interest, where the lower eye scatter may result in windscreen scatter effects becoming more prominent. For such experiments, it would be beneficial to match the irradiance levels at the front of the windscreens to allow direct comparison of threshold angles, without any subsequent curve fitting. It would also be advantageous to have scatter and transmission measurements at a range of wavelengths, potentially also using a spectrometer to cover the complete visible band. Improving the beam quality would be another target for such experiments, with the aim of achieving a more consistent and well-characterized beam across the entire windscreen. Studying the impact of windscreen scatter by itself would also be of interest, to investigate visual impacts when the laser is not incident upon the eye but still illuminates the windscreen.
Finally, the authors have recently issued a laser eye dazzle safety framework  that would benefit from consideration of windscreen effects. The framework provides a table of Maximum Dazzle Exposure (MDE) values covering different ambient light levels (night, dusk/dawn and day) and different dazzle levels (very low, low, medium and high). This present study suggests that windscreen scatter will have more of an effect at lower ambient light levels, so it is likely to have a bigger impact across the night (0.1 cd∙m−2) and dusk/dawn (10 cd∙m−2) conditions. It also indicates that the effect will be greater at larger angles, so it is likely to have a bigger impact across the medium (10° radius) and high (20° radius) dazzle levels. With the additional experimental work outlined above, it is anticipated that one or more additional MDE tables could be generated for applications involving windscreens. Such tables would need to accommodate a range of factors including the material type and the surface condition of the windscreen.
The authors wish to thank the Navy Medical Research Unit (NAMRU), Dayton, Ohio (Gary Labance, Matt Lee, William Becker, Dr Mike Reddix and Lt Cdr John Bradley) for the robot system used in the human observer experiments, and Dr. James R. Dykes for help with statistical analysis.
The opinions expressed on this document, electronic or otherwise, are solely those of the author(s). They do not represent an endorsement by or the views of the United States Air Force, the Department of Defense, or the United States Government.
DSTL/JA106052. Content includes material subject to © Crown copyright (2018), Dstl. This material is licensed under the terms of the Open Government Licence except where otherwise stated. To view this licence, visit http://www.nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: firstname.lastname@example.org.
References and links
1. LaserPointerSafety.com, “US: 6,753 laser/aircraft incidents reported to FAA in 2017; 9% lower than 2016,” http://www.laserpointersafety.com/news/news/other-news_files/d5fca5f3330bef52f1c5465b03502776-569.php#on.
2. Joint Non-Lethal Weapons Program, “Non-lethal optical distracters fact sheet”, (JNLWD, 2016), http://jnlwp.defense.gov/Portals/50/Documents/Press_Room/Fact_Sheets/NL_Optical_Distracters_Fact_Sheet_May_2016.pdf
4. C. A. Williamson, J. M. Rickman, D. A. Freeman, M. A. Manka, and L. N. McLin, “Measuring the contribution of atmospheric scatter to laser eye dazzle,” Appl. Opt. 54(25), 7567–7574 (2015). [CrossRef] [PubMed]
6. M. J. Allen, “Windscreen dirt and surface damage effects,” Aust. J. Optom. 58, 180–189 (1975).
7. P. Padmos, “Glare and tunnel entrance lighting: effects of stray-light from eye, atmosphere and windscreen,” CIE J. 3, 1–24 (1984).
8. P. A. Smith, “A study of the transient effects of high energy laser light on visual function,” PhD Thesis, University of London (1996).
9. P. L. Marasco and H. L. Task, “The effect on vision of light scatter from HMD visors and aircraft windscreens,” Proc. SPIE 3689, 2–10 (1999). [CrossRef]
10. P. L. Marasco and H. L. Task, “The measurement of visual performance through scattering visors and aerospace transparencies,” Proc. SPIE 4361, 188–197 (2001). [CrossRef]
11. D. W. Blick, J. M. Beer, W. D. Kosnik, S. Troxel, A. Toet, J. Walraven, and W. Mitchell, “Laser glare in the cockpit: psychophysical estimates versus model predictions of veiling luminance distribution,” Appl. Opt. 40(10), 1715–1726 (2001). [CrossRef] [PubMed]
12. A. Toet, J. K. IJspeert, H. Vos, and J. Walraven, “Computing the retinal image profile,” Tech.Rep. TNO-TM 1995C–035 (TNO Human Factors Research Institute, Soesterberg, Netherlands, 1995).
13. A. Toet and J. W. A. M. Alferdinck, “Effects of high power illuminators on vision through windscreens and driving behavior,” Proc. SPIE 8898, 88980I (2013). [CrossRef]
14. C. A. Williamson, L. N. McLin, M. A. Manka, J. M. Rickman, P. V. Garcia and P. A. Smith, Data to accompany paper: “The impact of windscreen scatter on laser eye dazzle,” (2018), https://doi.org/10.6084/m9.figshare.6755831.
15. C. A. Williamson, L. N. McLin, J. M. Rickman, M. A. Manka, P. V. Garcia, W. T. Kinerk, and P. A. Smith, “Wavelength and ambient luminance dependence of laser eye dazzle,” Appl. Opt. 56(29), 8135–8147 (2017). [CrossRef] [PubMed]
17. J. Vos, B. Cole, H.-W. Bodmann, E. Colombo, T. Takeuchi, and T. J. T. P. van den Berg, “CIE Equations for Disability Glare,” CIE TC Report CIE 146, 2002 (2002).
18. Air Force Medical Support Agency, “Protection of human subjects and adherence to ethical standards in Air Force supported research,” Air Force Instruction DODI3216.02_AFI 40–402, Air Force Medical Support Agency, 7700 Arlington Blvd Falls Church, VA 22042 (2014).
20. C. A. Williamson, “Simple computer visualization of laser eye dazzle,” J. Laser Appl. 28(1), 012003 (2016). [CrossRef]
21. G. S. Spencer, P. Shirley, K. Zimmerman, and D. P. Greenberg, “Physically-Based Glare Effects for Digital Images”, R. Cook, Ed., pp. 325–334, SIGGRAPH 95 Conference Proceedings, Annual Conference Series, ACM SIGGRAPH, Addison Wesley, (1995). [CrossRef]
22. J. Vos, “Disability glare – a state of the art report,” CIE J. 3, 39 (1984).
23. C. A. Williamson and L. N. McLin, “Nominal Ocular Dazzle Distance (NODD),” Appl. Opt. 54(7), 1564–1572 (2015). [CrossRef]
24. C. A. Williamson and L. N. McLin, “Determination of a laser eye dazzle safety framework,” J. Laser Appl. 30(3), 032010 (2018). [CrossRef]