Alterna · учебный материал
Cameras and technical vision
CMOS vs CCD sensors. Bayer filter. Dynamic range. Industrial cameras. 3D scans. Spectrum ranges.
Что в этом материале
- Key Camera Characteristics
- Spectre ranges
- 3D scanning
- Comparison of camera characteristics
- Mini-test: 5 questions
Содержание и основные понятия
By table of contents
Alterna · Chapter 20.3
Cameras and technical vision
Sensors, Bayer filter, dynamic range, industrial cameras, 3D scans.
As the camera sees the image
Digital camera converts light into an electrical signal through
sensor
Each sensor pixel measures the intensity of the light, but not the colour.
Bayer filter
.
CMOS (CMOS)
Active Pixel Sensor is a pixel that has its own transistor amplifier.
Energy efficiency
: consumes ~100 MW (vs CCD ~ 2 W). Faster reading (rolling shutter) is dominated by consumer chambers and industry.
CMOS energy efficiency 10 - 20 * higher than CMOC cameras can run from the battery for hours
♪ CCD (CCD)
The Charge-Cupled Vision is a pixel charge that is carried over the matrix to the edge of the force.
High image quality
: Low noise, high equilibrium, used in science chambers, astronomy, where accuracy is important, not energy consumption.
CCD is still used in scientific applications: astronomy, microscopics, spectrometry
Sensor pixel structure
Pixel (photosite) is a sequence of elements.
♪ Microlinsa
→
🎨 Color filter (Bayer)
→
☀️ Photodiode
→
⚡ Capacitor
→
📡 Readout circuit
Pixel size
is the size of the light-sensitive cell in micrometers (um)
Typical values: 0.8 um (smartphones), 1.4 - 2.4 um(s), 3.9 um [industrial]
Formula:
Large pixels ♪ more light ♪ less noise ♪ better with low lighting ♪
Example:
48 MP Pixel Sensor 0.8 μm vs 12 MP with 1.9 μm pixel
With ISO 3200:
12 MP Camera Better
- big pixels collect more light
💡
The path of light through the lens to the sensor
The pattern of beams passing through the camera lens
Incoming
rays
Elements of the lens
Refrigeration
RGB Pixel Sensor
R
G
B
G
R
G
Red beam R-suppixel ♪ Green beam ♪ G-sub-pixelle ♪ Blue beam B-subpixels
●
The red beam passes through the lens and hits the R-pixel.
●
Green beam (×2) is the maximum eye sensitivity
●
The blue beam passes through the Bayer filter
📷
Microlinsa focuses the light on the photodiode
⚡
Photodiode — As a Pixel of Camera "sees" Light
Photodiode working principle (sub-pixel element)
P-type
(positive carriers)
+
+
+
+
+
N-type
(negative carriers)
−
−
−
−
−
Poor zone
(depletion region)
Photon
+
−
+
−
+
−
3.3V
DC
A
Amperometer
Reverse
How does light turn into a signal?
1.
There are two layers in the pixel:
P-type
(with holes - places where electrons may be) and
N-type
(over electrons) between them is the boundary.
2.
When light (fotons) hits this border, it kicks the electrons out of the atoms.
3.
The electric field on the border divides them: electrons go one way, holes go the other way.
Electrical current
.
4.
The brighter the light, the more electron-hole pairs, the greater the current. By measuring this current, the camera determines
brightness and colour
The light on the pixel.
●
Animation: yellow points — photons of light moving towards p-n transition
●
Green particles — electrons running along chains (to create currents)
●
Purple area - P type (positive carriers), red - N type (negative)
⚡
The greater the light — the stronger the current — the brighter the pixel in the image
🎨
Bayer filter: How does one sensor color work?
Each sensor pixel is covered with a single coloured light filter (R, G or B). Green filters are twice as large as the eye is more sensitive to green. Demosaiking (interpolation) restores the full RGB for each pixels.
■ R
- Red filter only misses the red spectrum
■ G
- green (×2, half pixels) - maximum sensitivity
■ B
- blue filter
Patterne RGGB (normal) ♪ BGGR, GRB, CYGM (alternate)
Key Camera Characteristics
♪ Dynamic range (DR)
Maximum signal to noise ratio: DR = 20·log
10
(Full Well / Red Noise) [dB] Typical values: 60–80 dB (consumer), 90+ dB(scientific). HDR screen ~ 10,000:1 = ~80 db. The higher the detail in the shadows and lights.
Formula: DR (dB) ~ 6.02 × N + 1.76, where N is the Bacterium of ATP 12 bits ~ 75 dB.
ISO and noise
ISO
The higher the ISO is, the more bright the image is when the light is weak, but the noise is greater (grain).
Base ISO
===Sensor response===HDR mode: Several ISO and a tape.
Sony A7S III: ISO up to 409600. iPhone: Smart HDR with automatic brecking of exposure.
Megapixeli: real vs efficient
Effective MPs
===Reception===The film is actually used for filming.
More MPs
Not always better: the pixel becomes smaller than the noise in the light. 20x30 cm is enough to print an 8–12 MP. 50+ MP is for large prints or copier.
Canon EOS R5: 45 MPs, but only the central ~17 MP ×1.6 is used for the production of video 4K
♪ Machine Vision
Industrial automation cameras:
Area scan
(2D flat image) and
Line scan
Interfaces: USB3 Vision, Gige Vision, CoaXPress. Permits: VGA to 150 MP.
Line-scan camera: 4,096 pixels, 120 kHz lines/s ~ 500 million pixel/s for continuous tape
ISO 100
noise: 2e -
ISO 800
noise: 6e -
ISO 3200
noise: 18e -
ISO 12800
noise: 45e -
Spectre ranges
UV
10 - 400 nm
Purple.
blue
green
Red
Orange.
IC
700 - 1000 nm
Thermography
>1000 nm
Ultraviolet
Visible (380 - 700 nm)
Infrared (700 - 1000 nm)
Thermography (>1000 nm)
Application
UV:
Forensics, fluorescence, sterilization.
Vis:
It's just a picture, a auto-focused picture.
NIR(I):
Night vision, agriculture (NDVI).
Thermal:
Heat sensors, industrial controls, fire safety.
3D scanning
📡 Structured Light (Structured Light 3D)
The projector displays
Structured closet
The camera looks at the distortion of the pathtern, the stereo triangulation: the parallax calculates depth. Accuracy: up to 10 μm (laboratory systems).
Kinect v1: Structured Light. iPhone Face ID: Point Paternus of 30,000 IC Dots.
⏱️ Time of Flight (ToF)
The emitter sends the light pulse, measures
time of return
The resolution depends on the accuracy of the timer (~picoseconds). Modern ToF cameras: 640×480, accuracy ~1–5 mm at distances up to 5 m.
ToF is used in LiDAR (autonomous cars), AR (iPhone LiDar), industrial metrology.
Comparison of camera characteristics
Parameter
Smartphone (iPhone 15)
(Sony A7IV)
Industrial (Basler acA4096)
Heat Vigilance (FLIR T640)
Sensor type
CMOS BSI
CMOS Exmor R
CMOS
Microbolometer
Authorization
48 MP (12 MP effects).
33 MP
12 MP
640×480
Pixel size
0.8 μm
5.1 um
3.45 μm
17 μm
ISO range
100–25600
100–51200
—
—
Bit depth
10 - 12 bits
14 bits RAW
8 - 12 bits
14 bits
Dynamic range
~12 EV
~15 EV
~60 dB
~70 dB
Interface
Internal
PCIe / Wi-Fi
USB3 / GigE
USB3 / HDMI
Frequency of staff
60 fps (4K)
30 fps (4K)
30 fps
30 fps
Price
of the United Nations
~200 000 ₽
~80 000 ₽
~500 000 ₽
🎯
Mini-test: 5 questions
XP: 0