关于「防护与颜色」,《黑视素光谱、褪黑素抑制与防蓝光镜片》给出的定量结论是什么?

结论:第五章描出防护与显色的前沿:满足显色保真指数 70 的最佳长通镜片只防护 0.364,阻挡 440 至 514 nm 的陷波镜片以 70.3 达到 0.633。

数据口径:书中的光谱、镜片、人与日程均为示意模型;全部计算由书中程序生成;本书不是医疗建议,不诊断也不治疗任何疾病,夜间驾驶时绝不可佩戴有色镜片。 书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 0.364, 0.633, 440, 70, 70.3

   P = melanopic protection, F = lens fidelity index
   LED warm 3000K: 29 of 29 edges on the front
     best long-pass with F >= 70: edge 480 nm, P 0.364, F 71.5
     long-pass edge 485 nm: P 0.395, F 68.8
     long-pass edge 500 nm: P 0.510, F 62.4
     long-pass edge 510 nm: P 0.601, F 58.6
     best notch with F >= 70: blocks 440-514 nm, P 0.633, F 70.3
   phone OLED day: 29 of 29 edges on the front
     best long-pass with F >= 70: edge 470 nm, P 0.407, F 71.9
     long-pass edge 485 nm: P 0.528, F 51.5
     long-pass edge 500 nm: P 0.587, F 39.7
     long-pass edge 510 nm: P 0.646, F 38.3
     best notch with F >= 70: blocks 445-479 nm, P 0.514, F 70.2
查看完整验算程序
# Chapter 5. p10: The Pareto front between protection and colour. A
# family of long-pass lenses T(l) = 0.02 + 0.90 / (1 + exp(-(l - c)/6))
# with the edge c moved from 420 nm to 560 nm. For each source we keep
# the edges that no other edge beats on both protection and fidelity,
# report the best protection with the fidelity index at least 70, and
# the lenses with edges at 485, 500 and 510 nm. Then we try a second
# family, notch filters that block a band [a, b) and pass both sides,
# and report the best notch with fidelity at least 70.
import math
from math import exp, expm1

L = range(380, 781)                     # wavelength grid, 1 nm steps


def g(x, m, s1, s2):                    # two-sided Gaussian (Wyman 2013)
    return exp(-0.5 * ((x - m) / (s1 if x < m else s2)) ** 2)


def vlam(l):                            # photopic V(lambda)
    return 0.821 * g(l, 568.8, 46.9, 40.5) + 0.286 * g(l, 530.9, 16.3, 31.1)


def opsin(l, lm=480.0):                 # Govardovskii A1 template
    x = lm / l
    a = 0.8795 + 0.0459 * exp(-(lm - 300) ** 2 / 11940)
    al = 1 / (exp(69.7 * (a - x)) + exp(28 * (0.922 - x))
              + exp(-14.9 * (1.104 - x)) + 0.674)
    lb, bb = 189 + 0.315 * lm, 0.195 * lm - 40.5
    return al + 0.26 * exp(-((l - lb) / bb) ** 2)


def lens_od(l):                         # crystalline lens density, approx.
    return 1.76 * exp(-(l - 400) / 28.0)


RAW = {l: opsin(l) * 10 ** -lens_od(l) for l in L}
SMEL = {l: RAW[l] / max(RAW.values()) for l in L}
K_D65 = 1.3262 / 1000                   # W/lm, melanopic ELR of D65


def gs(l, m, f):                        # Gaussian band: peak m, FWHM f
    return exp(-2.7726 * ((l - m) / f) ** 2)


def planck(l, t):
    m = l * 1e-9
    return 1.0 / (m ** 5 * expm1(1.4388e-2 / (m * t)))


DEVICES = {                             # relative spectral power
    "phone OLED day": lambda l: (1.12 * gs(l, 460, 25) + gs(l, 530, 35)
                                 + 1.05 * gs(l, 620, 30)),
    "phone night mode": lambda l: (0.44 * gs(l, 460, 25)
                                   + gs(l, 530, 35)
                                   + 1.80 * gs(l, 620, 30)),
    "LED cool 6500K": lambda l: (2.32 * gs(l, 450, 20) + gs(l, 545, 100)
                                 + 0.50 * gs(l, 630, 80)),
    "LED warm 3000K": lambda l: (0.96 * gs(l, 450, 20) + gs(l, 555, 110)
                                 + 1.30 * gs(l, 625, 90)),
    "incandescent": lambda l: planck(l, 2856),
}


def der(spec, t=lambda l: 1.0):         # mel EDI per lux, through t
    em = sum(spec(l) * t(l) * SMEL[l] for l in L)
    ev = 683.0 * sum(spec(l) * t(l) * vlam(l) for l in L)
    return em / K_D65 / ev


def mel_through(spec, t):               # melanopic share a filter passes
    return (math.fsum(spec(l) * t(l) * SMEL[l] for l in L)
            / math.fsum(spec(l) * SMEL[l] for l in L))


def xbar(l):
    return (1.056 * g(l, 599.8, 37.9, 31.0) + 0.362 * g(l, 442.0, 16.0, 26.7)
            - 0.065 * g(l, 501.1, 20.4, 26.2))


def zbar(l):
    return 1.217 * g(l, 437.0, 11.8, 36.0) + 0.681 * g(l, 459.0, 26.0, 13.8)


SAMPLES = [lambda l, m=m: 0.2 + 0.5 * exp(-((l - m) / 50.0) ** 2)
           for m in (430, 470, 500, 530, 560, 590, 620, 660)]


def uvY(spec, refl, k):
    x_, y_, z_ = (k * math.fsum(spec(l) * refl(l) * f(l) for l in L)
                  for f in (xbar, vlam, zbar))
    q = x_ + 15 * y_ + 3 * z_
    return 4 * x_ / q, 6 * y_ / q, y_


def cd(u, v):
    return (4 - u - 10 * v) / v, (1.708 * v + 0.404 - 1.481 * u) / v


def fidelity(spec, t):
    """CIE 13.3 arithmetic; reference = the bare source, test = lens."""
    ref, tst = spec, (lambda l: spec(l) * t(l))
    kr = 100 / math.fsum(ref(l) * vlam(l) for l in L)
    kt = 100 / math.fsum(tst(l) * vlam(l) for l in L)
    ur, vr, _ = uvY(ref, lambda l: 1.0, kr)
    uk, vk, _ = uvY(tst, lambda l: 1.0, kt)
    (cr, dr), (ck, dk) = cd(ur, vr), cd(uk, vk)
    out = []
    for s in SAMPLES:
        u1, v1, y1 = uvY(ref, s, kr)
        u2, v2, y2 = uvY(tst, s, kt)
        c2, d2 = cd(u2, v2)
        den = 16.518 + 1.481 * cr / ck * c2 - dr / dk * d2
        u2 = (10.872 + 0.404 * cr / ck * c2 - 4 * dr / dk * d2) / den
        v2 = 5.520 / den
        w1, w2 = 25 * y1 ** (1 / 3) - 17, 25 * y2 ** (1 / 3) - 17
        de = ((13 * w1 * (u1 - ur) - 13 * w2 * (u2 - ur)) ** 2
              + (13 * w1 * (v1 - vr) - 13 * w2 * (v2 - vr)) ** 2
              + (w1 - w2) ** 2) ** 0.5
        out.append(100 - 4.6 * de)
    return math.fsum(out) / len(out)


def sig(l, c, w=6.0, lo=0.02, hi=0.92):
    return lo + (hi - lo) / (1.0 + exp(-(l - c) / w))


def sweep(spec, edges=range(420, 565, 5)):
    rows = []
    for c in edges:
        t = (lambda l, c=c: sig(l, c))
        rows.append((c, 1 - mel_through(spec, t), fidelity(spec, t)))
    return rows


def best_notch(spec, floor=70.0):
    best = None
    for a in range(430, 520, 5):
        for b in range(a + 10, 560, 5):
            t = (lambda l, a=a, b=b: 0.02 if a <= l < b else 0.92)
            p, f = 1 - mel_through(spec, t), fidelity(spec, t)
            if f >= floor and (best is None or p > best[0]):
                best = (p, f, a, b)
    return best


def pareto(rows):
    return [r for r in rows if not any(
        o[1] >= r[1] and o[2] >= r[2] and o != r for o in rows)]


if __name__ == "__main__":
    print("   P = melanopic protection, F = lens fidelity index")
    for name in ("LED warm 3000K", "phone OLED day"):
        rows = sweep(DEVICES[name])
        front = pareto(rows)
        ok = [r for r in rows if r[2] >= 70]
        best = max(ok, key=lambda r: r[1])
        print("   {}: {} of {} edges on the front".format(
            name, len(front), len(rows)))
        print("     best long-pass with F >= 70: edge {} nm, P {:.3f},"
              " F {:.1f}".format(*best))
        for r in rows:
            if r[0] in (485, 500, 510):
                print("     long-pass edge {} nm: P {:.3f}, F {:.1f}".format(
                    *r))
        p, f, a, b = best_notch(DEVICES[name])
        print("     best notch with F >= 70: blocks {}-{} nm, P {:.3f},"
              " F {:.1f}".format(a, b - 1, p, f))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《黑视蛋白光谱积分、蓝光褪黑素抑制与镜片透射率优化》第 5 章「防护与颜色」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-104 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026104,
  author    = {Fan, Yafu},
  title     = {Melanopic Spectral Integration, Blue-Light Melatonin Suppression and Lens Transmittance Optimization},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论