www.lesswrong.com/posts/SzjfsWe9bfHSX8fRg/petapixel-cameras-won-t-exist-soon#com...
2 corrections found
Commercial satellite footage is not allowed below 10 cm resolution that enables facial recognition.
A ground sample distance of 10 cm per pixel is far too low-resolution to enable facial identification/recognition; standards and industry guidance typically require on the order of tens of pixels across a face (hundreds of pixels per meter).
Full reasoning
The post claims that 10 cm resolution “enables facial recognition.” But 10 cm/pixel implies only ~1–2 pixels across a human face.
- Axis Communications (citing the European CCTV standard EN50132-7:2012) states that a 16 cm-wide face should cover at least 40 pixels (and Axis suggests 80 pixels or more in challenging conditions) for identification. (axis.com)
- 40 pixels across 16 cm corresponds to ~4 mm/pixel (0.004 m/pixel).
- Axis also expresses identification needs as ~250 pixels per meter (and 500 px/m in challenging conditions). (axis.com)
- 10 cm/pixel equals 10 pixels per meter, which is ~25× lower pixel density than 250 px/m.
Given these widely used identification requirements, “10 cm resolution” does not enable facial identification/recognition in the ordinary meaning used in surveillance and forensics contexts, so the sentence as written is incorrect (regardless of what the legal distribution limit is in any particular country).
2 sources
- Axis Communications — Identification and Recognition (PDF)
“Based on older CCTV recommendations, a 16 cm-wide face should cover at least 40 pixels … However, Axis suggests 80 pixels or more for identification in challenging conditions.”
- Axis Communications — Panoramic cameras (White paper, PDF)
“Axis recommends 25 px/m … 125 px/m … and 250 px/m (76 px/ft) to identify someone. … Identification (challenging conditions) 500 px/m.”
Gigapixel cameras allow you to do facial recognition (0.1 metre resolution) from a distance of 1-10 kilometres.
0.1 m (10 cm) per-pixel resolution is far too coarse for facial identification; guidance commonly requires about 40+ pixels across a face (hundreds of pixels per meter), not ~1–2 pixels.
Full reasoning
The key factual error is equating 0.1 m resolution with being sufficient for facial recognition/identification.
- Axis Communications notes (referencing EN50132-7:2012) that a 16 cm-wide face should cover at least 40 pixels for identification (and suggests 80+ pixels in challenging conditions). (axis.com)
- If each pixel is 0.1 m (10 cm), then a 0.16 m face spans only 0.16/0.1 = 1.6 pixels across—nowhere near 40–80 pixels.
- Axis also summarizes identification needs in terms of pixel density on the subject, recommending 250 px/m for identification in good conditions (500 px/m in challenging conditions). (axis.com)
- 0.1 m/pixel corresponds to 10 px/m, far below 250 px/m.
So while gigapixel imaging can capture large scenes at high detail, the statement that facial recognition is achievable at 0.1 m resolution is contradicted by established identification-resolution guidance/standards.
2 sources
- Axis Communications — Identification and Recognition (PDF)
“Based on older CCTV recommendations, a 16 cm-wide face should cover at least 40 pixels … However, Axis suggests 80 pixels or more for identification in challenging conditions.”
- Axis Communications — Panoramic cameras (White paper, PDF)
“The level of detail depends on the pixel density … Axis recommends … 250 px/m … to identify someone.”