Visual Performance Models in Road Lighting: A Historical Perspective

ABSTRACT

The use of visual performance models in road lighting is an old idea: it was first proposed in the 1930s by Waldram with the Revealing Power, and by Roper and Howard who used the notion of visibility distance. The Visibility Level (VL) concept was then proposed by Blackwell, and the Relative Visual Performance by Rea and Ouelette. At the turn of the 21th century, some standards have considered using the VL in order to rate lighting installations through the Small Target Visibility concept. However, the use of visual performance indexes in lighting standards was recently withdrawed, which raises a question: what happened?

KEYWORDS:

• Visual performance
• visibility
• target detection
• road lighting

Acknowledgments

I wish to thank Céline Villa, Eric Dumont & Jean-Philippe Tarel for their careful reading of the manuscript and fruitful discussions, as well as the anonymous reviewers for their helpful advice.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. With this in mind, I will not go into the details of computational formulas.

2. Formerly Transactions of the IES (London), and the Journal of the IES.

3. Speed control is sometimes included in the first level.

4. Lighting engineering and vision science only have weak links. Among others, Mark Rea suggested that lighting engineers should have a better understanding of vision science concepts (Rea 1982). His main idea was to use, as a bridge between the two communities, a concept familiar to both: visual performance.

5. The quantification of glare was already taken into account, thanks to the recent work by Holladay (1926).

6. The idea of milestones was taken from Wilson (1942).

7. Wood (1936) explains that during the Great Depression, several road authorities decided to switch off the road lighting on their network to save money. A few years later, it was swiched on again, because of nighttime accidents.

8. The “ simplest imaginable”, or basic task: any change in the task makes it more difficult.

9. The contrast threshold for 90% or 99% of target detection can be computed from the SD ($\sigma$), which depends on the experimental conditions. Considering that $\sigma$ is roughly constant, Adrian (1989) computed a constant factor to convert the 50% threshold to a nearly 100% threshold (see Section 3.2).

10. The adaptation luminance was, in this case, also the background luminance.

11. H. R. Blackwell was the chairman of the Technical Committees which produced the CIE reports 19 and 19.2.

12. From the same comparison, Rea focused on response time alone (see Section 3.3).

13. This model is in line with the main signal processing theory at the time (Shannon 1948), including the ideas of a signal modulation on the transmitter side, and a modulation sensitivity on the receiver side.

14. Among these factors, Blackwell (1946) intensively explored the “four factors” highlighted by Cobb and Moss (1928) (see Section 2), while Lamar et al. (1947) addressed the target shape. The luminance uniformity around the target was studied by Finch (1959) and Chorlton and Davidson (1959), the colored contrast by Eastman (1968a). The effect of glare was described by Holladay (1927) and Stiles (1929a, 1929b), while Boynton et al. (1970) addressed transient glare.

15. Three levels of performance are described: the visibility in the usual sense, which corresponds to optimal viewing conditions; the visibility when visual search is taken into account; and the true visual performance, which also depends on individual factors (training, fatigue, motivation, etc.).

16. The detection of a disc 4ʹ in diameter, displayed 200 ms in central vision.

17. In the next CIE report, the dependence of the RCS on the target size and on the observer’s age was considered (CIE 1981a), see section 3.1.3.

18. It is recommanded to estimate a visual performance in two steps, in an integrating sphere: first in the Reference condition, then in the modified conditions (glare, etc.) with a visibility-meter. Only if this two-step approach is not convenient (for instance, on the road) it is proposed to consider the VL in the real task conditions.

19. Be careful that these words also denote Rea’s model (Rea 1986).

20. The study from Gallagher and Meguire (1975) is reported in Sec. 3.4.1; Another one, with 8 drivers, repeated the same drive with and without optical filters (Economopoulos 1978).

21. Cf. section 3.4.2.

22. The angular size of the target was 10ʹ, with contrasts between 0.2 and 0.3. Based on oculomotor data collected in driving situations, the observation time was set to 0.2 s (Narisada and Yoshikawa 1974; Zwahlen 1985).

23. Above some critical angle ${\alpha }_c$, Weber’s law applies and the detection threshold $\mathrm{\Delta }{L}_{th}$ does not depend on the target size; for smaller targets, ${\alpha }^2\mathrm{\Delta }{L}_{th}$ is constant (Ricco’s law).

24. This issue was first raised by Hills (1976).

25. VL can be written either as $\mathrm{\Delta }L/\mathrm{\Delta }{L}_{th}$ or as $C/C_{th}$.

26. For instance, a minimal VA is required for a driving license.

27. This value is consistent with Gallagher and Meguire (1975): above VL = 15, the drivers had no collision with the obstacles (see section 3.4.1).

28. In previous studies, the contrast polarity was considered negligible (Blackwell 1946; Herrick 1956; Hills 1976; Judd and Eastman 1971).

29. Three formulas are proposed for $\mathrm{\Delta }{L}_{\mathrm{\infty }}$ and $\varphi$, depending on the range of $L_b$.

30. This is true for tasks with very few errors.

31. From the classical Naka-Rushton curve, which describes the response of a biological system as a function of the stimulation (Naka and Rushton 1966).

32. It was also realized that the motor response time is uncompressible, leading to a minor modification of the formula.

33. The Contrast Sensitivity Function (CSF) was considered too complex, and Hills preferred this “single-channel” model (see section 6.2). Other such models were available at that time: Blackwell and Smith (1958) used corrective factors, while Fry (1947) proposed to use sensitivity data from rectangles rather than disc targets. Beurle et al. (1968) proposed a two-step model: the first one has a central symetry, while the second uses a line detector.

34. The absolute threshold ($L_b=0$) was first modeled (Fry 1965; Nolan 1957); Beurle et al. (1968) extended the formula to $L_b>0$.

35. The dependence of the critical angle $r_m$ on $L_b$ is taken from psychophysical data (Beurle et al. 1968; Blackwell 1946).

36. Although Hill’s model was published in Lighting Research and Technology, it was relatively unnoticed by lighting practicionners, or even by the CIE and the IESNA.

37. It is well known, for instance, that the target size impacts the detection distance.

38. Due to the inter-individual variability, all correlations were weak. The higher value for R${ }^2$ was 0.3.

39. The experimenter drove at constant speed.

40. Ménard and Cariou (1994) collected subjective estimates of the visibility, just like Dunbar (1938).

41. In CIE report n${ }^{\circ }$115, using the luminance criterion, a longitudinal uniformity $U_l=0.7$ is needed, while using the STV, $U_l=0.2$ is considered enough.

42. Kokoschka (1985) considered the visibility of a heterogeneous target (CIE 1992a). He derived a formula to compute its visibility from the CIE model in one specific case, with homogeneous sub-targets of equal areas.

43. The luminance values on the sphere were computed, not measured, because measurements on a sphere is not easy.

44. Lecocq (2000) also tested his idea on the reduced scale (1:15) model of a 225-m road section with road lighting; the small luminaires could be tuned in various configurations.

45. With a uniform background, these three estimates are equal.

46. This experimental design is similar to Lecocq’s (Lecocq 2000).

47. This task was selected because is was considered close to driving in terms of the perceptive and cognitive processes involved.

48. In the dynamic condition, the vehicle speed was fixed.

49. Cars, pedestrians, road signs, and a STV “small target”; some were colored and textured, some were gray and uniform.

50. Remember that the first version of the RVP used a typical interior lighting task as the reference task (the numerical verification task).

51. CIE report n${ }^{\circ }$145 describes two types of visual tasks, needing two types of visual performance models, because the visual pathways and the underlying physiological mechanisms are different. In an acuity task, and each time the vision of details is involved, the signal processing follows the parvo-cellular visual pathway of the brain; in simple detection dasks, the magno-cellular pathway is more involved. This may be the physiological mechanism explaining that the performance models are not the same with stimuli of high and low spatial frequencies.

52. The mean VL value on a grid was not broadly accepted, see above Sec. 4.1.

53. TC: Technical Committee.

54. This TC on Visibility Design for Roadway Lighting did not publish any report so far.

55. Remember that Cobb and Moss (1928) already used gratings as their reference stimuli. But at that time, their benefits were not fully understood.

56. The good point with linear filters is that one can use the linear systems theory (Wandell 1995).

57. The number of spatial scales is estimated between 4 and 6 (Wilson and Bergen 1979; Wilson et al. 1983).

58. In peripheral vision, the surface of the visual cortex associated to a given solid angle decreases with eccentricity, which can be described with a cortical magnification factor (Cowey and Rolls 1974). Overington (1982) used this factor to estimate the visual performance outside the fovea.

59. In other words, a task would be referred to as “complex” if it needs high spatial frequency information.

60. This includes some uncontrolled factors, such as the gaze direction, but there is a growing literature about the range of gaze in driving situations (Fotios et al. 2015; Foulsham et al. 2011; Land and Lee 1994; Lappi 2014).

61. Apart from visual performance, there is a growing concern about the unwanted effects of lighting. In addition to disability glare, this includes discomfort glare, blue light hazard, circadian rythm and the protection of biodiversity.

62. Remember that the observers in Blackwell’s studies all had a very good vision.

63. The CIE has recently highlighted the 10 main topics of a Research Strategy for the coming years, among which: Metrology for Advanced Photometric and Radiometric Devices.

Additional information

Funding

There is no funding source to report.