【钠瞳 Sodium Pupil】四：模型与方法

四、结果

气象条件	策略	事件捕获率	全天耗电	停止工作时间	识别次数
常规冬季日	常规相机	56.1%	29.5 Wh	8.8 h	0
常规冬季日	定时阈值	0.0%	26.4 Wh	18.4 h	0
常规冬季日	钠瞳	92.5%	13.8 Wh	存活 24 h	66
极寒低日照	钠瞳	93.1%	15.2 Wh	存活 24 h	71
连续阴雪	钠瞳	89.5%	22.4 Wh	存活 24 h	131

日照分别为 6.27 / 2.63 / 1.02 Wh；数值为仿真假设值，非第三方实测。

几个值得单独看一眼的点：

• 常规相机 拍了 1485 张，却在 8.8 小时就断电——夜里等于没装；
• 定时阈值策略 当天一次事件都没认出来，因为它的判断依据里根本没有"事件"，只有"时间"和"电量"。它在 18.4 小时关机后，需要等人上山复位；
• 连续阴雪条件下，钠瞳主动把"认"降级为"看"，耗电从 13.8 Wh 升到 22.4 Wh，但仍保住 89.5% 的捕获率。

IV. Results

Scenario	Strategy	Event capture rate	Daily energy use	Time of shutdown	Identifications
Normal winter day	Conventional camera	56.1%	29.5 Wh	8.8 h	0
Normal winter day	Fixed threshold	0.0%	26.4 Wh	18.4 h	0
Normal winter day	Sodium Pupil	92.5%	13.8 Wh	Survives 24 h	66
Extreme cold, low solar	Sodium Pupil	93.1%	15.2 Wh	Survives 24 h	71
Continuous overcast snow	Sodium Pupil	89.5%	22.4 Wh	Survives 24 h	131

Daily solar energy is 6.27 / 2.63 / 1.02 Wh respectively. All figures are simulation assumptions, not third-party measurements.

A few points deserve a closer look:

• The conventional camera took 1485 photos but lost power after 8.8 hours — at night it was effectively not installed;
• The fixed-threshold strategy did not identify a single event that day, because its criteria contain no notion of "event" at all, only "time" and "charge". After shutting down at 18.4 hours it needs someone to climb the mountain and reset it;
• Under continuous overcast snow, Sodium Pupil voluntarily degraded "identify" to "look"; its consumption rose from 13.8 Wh to 22.4 Wh, yet it still preserved a capture rate of 89.5%.