关于「实战工具与检查器规格:周度远距辨识台账与预注册检验」,《服饰图案远距辨识与再认》给出的定量结论是什么?

结论:第八章交出蒸馏小工具「服饰图腾远距辨识与曝光阈值试算器」、按四件套分组的检查器规格、周度台账与预注册检验,单款检验约需 69 名匿名受访者。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。

数据口径:书中参数为示意,全部计算由书中程序生成。 本书不是商标或著作权法律意见,示意图案全部为作者自拟几何形,不点名评价任何真实品牌,不写任何潜意识影响宣称,街头测试只记录匿名对错,不采集人脸与个人资料。

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: 69

A. per-arm n, two designs at 10 m
0.40 vs 0.55: n =173 per arm,14 sessions
0.45 vs 0.60: n =173 per arm,14 sessions
0.60 vs 0.75: n =152 per arm,13 sessions
0.30 vs 0.50: n =93 per arm,8 sessions
B. exact one-sided test of recognition above a floor
p0 0.50, p1 0.65: n =69, reject at >=42, size 0.0456, power 0.8021
p0 0.30, p1 0.45: n =67, reject at >=27, size 0.0466, power 0.8147
p0 0.20, p1 0.35: n =56, reject at >=17, size 0.0432, power 0.8064
查看完整验算程序
# Chapter 8. p82: how many anonymous answers a street test needs.
# A: two designs at 10 m, recognition p1 vs p2, two-sided alpha 0.05,
#    power 0.80, normal approximation for two proportions.
# B: one design against a pre-registered floor p0 (one-sided exact
#    binomial test, alpha 0.05): smallest n whose exact power at the
#    hoped-for p1 reaches 0.80, found by enumeration.
# C: sessions needed at 25 anonymous answers per session (no faces, no
#    personal data are recorded; only right / wrong per distance).
import math
from statistics import NormalDist

Z = NormalDist()


def n_two(p1, p2, alpha=0.05, power=0.80):
    za, zb = Z.inv_cdf(1 - alpha / 2), Z.inv_cdf(power)
    pb = (p1 + p2) / 2.0
    num = (za * math.sqrt(2 * pb * (1 - pb))
           + zb * math.sqrt(p1 * (1 - p1) + p2 * (1 - p2))) ** 2
    return math.ceil(num / (p1 - p2) ** 2)


def tail(n, k, p):
    terms = [math.comb(n, j) * p ** j * (1 - p) ** (n - j)
             for j in range(k, n + 1)]
    return math.fsum(terms)


def n_exact(p0, p1, alpha=0.05, power=0.80):
    for n in range(10, 2000):
        k = n
        while k > 0 and tail(n, k - 1, p0) <= alpha:
            k -= 1
        if tail(n, k, p1) >= power:
            return n, k, tail(n, k, p0), tail(n, k, p1)
    return None


if __name__ == "__main__":
    print("A. per-arm n, two designs at 10 m")
    for p1, p2 in ((0.40, 0.55), (0.45, 0.60), (0.60, 0.75), (0.30, 0.50)):
        n = n_two(p1, p2)
        print("  {:.2f} vs {:.2f}: n = {:>4} per arm, {:>3} sessions".format(
            p1, p2, n, math.ceil(2 * n / 25)))
    print("B. exact one-sided test of recognition above a floor")
    for p0, p1 in ((0.50, 0.65), (0.30, 0.45), (0.20, 0.35)):
        n, k, a, pw = n_exact(p0, p1)
        print("  p0 {:.2f}, p1 {:.2f}: n = {:>3}, reject at >= {:>3},"
              " size {:.4f}, power {:.4f}".format(p0, p1, n, k, a, pw))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《服饰图腾完形视知觉捕获、远距高对比辨识与街头视觉模因驻留母典》第 8 章「实战工具与检查器规格:周度远距辨识台账与预注册检验」(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-390 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026390,
  author    = {Fan, Yafu},
  title     = {Apparel Graphic Gestalt Capture, Long-Range Contrast Legibility and Streetwear Visual Memetic Retention},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论