Robot Vision SS 2005 Matthias Rüther 1 ROBOT VISION Lesson 5: Camera Hardware and Technology...
-
Upload
jarrett-macy -
Category
Documents
-
view
218 -
download
0
Transcript of Robot Vision SS 2005 Matthias Rüther 1 ROBOT VISION Lesson 5: Camera Hardware and Technology...
Robot Vision SS 2005 Matthias Rüther 1
ROBOT VISION Lesson 5: Camera Hardware and Technology
Matthias Rüther
Robot Vision SS 2005 Matthias Rüther 2
Content
Camera Hardware– Sensors
– Video Data Transfer
– Mechanics
Optics– Lenses
– Macroscopic
– Telecentric
– Microscopic
Illumination– Illumination systems
– Mechanical Arrays
Robot Vision SS 2005 Matthias Rüther 3
Sensors
Goal: convert light intensity to electrical signal– Mostly visible light spectrum (~700nm to ~400nm)
provides color information, light intensity, like human eye
– Near infrared (~700nm to 5m)Similar properties as visible light, NO heat information; black sky, plants are
white, used for vegetation inspection, remote sensing, to detect reflective markers
Robot Vision SS 2005 Matthias Rüther 4
Sensors
– Ultraviolet (~400nm to ~240nm)Used with special illumination,
UV microscopy (resolution up to 100nm)
surface inspection (detecting cracks, fluid leaks etc.)
flame inspection (alcohol flames are barely visible to human eye)
Forensics (finger print, blood, etc.)
Robot Vision SS 2005 Matthias Rüther 5
Sensors
2 Basic Technologies– Charge Coupled Device– CMOS Sensor
Both are pixelated metal oxide semiconducters Accumulate in each pixel signal charge proportional to local
illumination intensity => spatial sampling function Properties
– Responsivity: amount of output signal per unit of input optical energy– Dynamic range: ratio of saturation level to signal threshold– Uniformity: consistency of response– Shuttering: start and stop of exposure– Speed: frame rate / readout time– Windowing: can subwindows of the chip be sampled?– Antiblooming– Biasing / Clocking– Reliability– Cost
Robot Vision SS 2005 Matthias Rüther 6
Charge Coupled Device
Robot Vision SS 2005 Matthias Rüther 7
CMOS Sensor
Robot Vision SS 2005 Matthias Rüther 8
CCD vs CMOS
Robot Vision SS 2005 Matthias Rüther 9
Line Sensor
Robot Vision SS 2005 Matthias Rüther 10
Line Sensor
Robot Vision SS 2005 Matthias Rüther 11
Video Data Transfer
Transfer of image data from Camera to System Memory
Properties:– Transfer distance
– Bandwidth / Framerate
– Analog / Digital
– Environment
– Cost
Robot Vision SS 2005 Matthias Rüther 12
Analog Video Signal
Video Standards– Composite Video: de facto standard for consumer products, combines
color, brightness and synchronisation data to one „composite“ signal.
– S-Video: Y/C or Component Video; splits video data in two channels: luminance (Y) and chrominance (C). Provides less granularity and sharper image. C = U/V for PAL and C = I/Q for NTSC
– RGB: standard for computer monitors. Four signals (red, green, blue, sync)
– Large distances possible (~20m). Higher frequencies degrade with length (low pass) and noise adds to the signal.
Robot Vision SS 2005 Matthias Rüther 13
Analog Video Signal
Broadcast Standards (combine technical and legal definitions)
– NTSC: National Television Standards Comitee; North/Central America, Mexico, Canada, Japan.
– PAL: Phase Alteration Line; UK, Western Europe, Middle East, Parts of Africa and South America
– SECAM: Systeme Electronic Pour Coleur Avec Memoire, similar to PAL, chrominance is FM modulated; France, Russia, parts of Africa, Eastern Europe
Format Country ModeSignal Name
Frame Rate [fps]
Vert. Line Resolution
Line Rate [lps]
Img. Size
NTSC US, Japan Mono RS-170 30 525 15.750 640x480
Color NTSC Color 29.97 525 15.734
PAL Europe except France
Mono CCIR 25 405 10.125 768x576
Color PAL Color 25 625 15.625
SECAM France, East.
Europe
Mono 25 819 20.475 N/A
Color 25 625 15.625
Robot Vision SS 2005 Matthias Rüther 14
Analog Video Signal
Scanning Process
Robot Vision SS 2005 Matthias Rüther 15
CameraLink
Serial Interface for digital image transfer. Standardized Fast (up to 2.38 Gbps) Not a High Volume Product -> expensive Max 10m cable, no power provided
Physical Layer: Low Voltage Differential Signaling (LVDS); high-speed, low-power general purpose interface standard; known as ANSI/TIA/EIA-644, approved in March 1996.– 350 mV nominal signal swing– Theoretical 1.923 Gbps
Connection Channellink: developed by Natioan Semiconducturs for flat panel displays, – 28bit I/O, serialized 7:1 and transferred– Up to 2.38 Gbps
Cameralink specializes Channellink for video data transfer.
Robot Vision SS 2005 Matthias Rüther 16
CameraLink
Robot Vision SS 2005 Matthias Rüther 17
IEEE 1394 (Firewire)
De-facto industrial standard
– Moderate volume product (Industrial cameras, Video Cameras, Webcams)
– Consists of both hardware and software specification– Completely digital--no conversion to analog – Data rates of 100, 200, or 400 Mb per second (800Mbps by 1394b)– Flexible--supports daisy-chain and branching cable configurations– Inexpensive – Max 4.5m cable length– Power provided by bus
– Invented by Apple in mid 90‘s as LAN bus (100Mbps)– Development hampered by license fees in 1998 ($1 per port)– Since 1999 owned by 1394LA ($0.25 per unit)– Firewire remains trademark of apple.
Robot Vision SS 2005 Matthias Rüther 18
USB 2.0
Upcoming rival for IEEE1394– Fast (480Mbps)
– High volume (available on every PC)
– Plug and Play
– Emerged from USB 1.1 (1995)
– Provides Power
– 5m cable length
– Master-Slave Architecture (IEEE1394: Peer to Peer)
– IEEE1394 is faster (10-70%), due to protocol architecture!
Robot Vision SS 2005 Matthias Rüther 19
Mechanics
Industrial cameras need to be ruggedized
– Up to 90% humidity– -5 to +50 degrees Celsius– Harder requirements for
outdoor/surveillance cameras
Common Sensor dimensions:– ¼“– 1/3“– ½“– 2/3“– 1“
Mounting usually by ¼“ screws Lens mount standards: C-mount and
CS-mount; 1“ thread; differing by flange focal distance
Robot Vision SS 2005 Matthias Rüther 20
Optics
… or how to calculate the focal length.
Lenses (or lens systems, a „compound“ lens) are used to project light rays on an image sensor.
If all rays originating from a distinct point of light intersect in one point on the image plane, a sharp image of this point is acquired.
Robot Vision SS 2005 Matthias Rüther 21
Lens Parameters
Magnification = size of image / size of object– E.g. size of object = 5cm; size of image =
5mm -> magnification = 0.1
– Depends on working distance (lens – object distance) -> impractical for standard lenses
Focal length = working distance * size of image / (size of object + size of image)– E.g. to capture a 1000m wide object from
500m on a CCD chip measuring 4.8x6.4mm, you need 3.2mm of focal length
Robot Vision SS 2005 Matthias Rüther 22
Lens Iris
The Iris limits the amount of light getting through the lens.
-> the image appears darker (avoids overexposure in bright scenes)
-> less lens area is used -> fewer lens errors are incorporated
-> sharpness is increased
Sharpness: theoretically impossible to focus 3D object, but:
– Blurred points of some size appear sharp to human eye (e.g. on 35mm film, 1/30mm spots appear sharp)
– -> „Depth of field“– In practice: max. blurred spot is 1
pixel
Robot Vision SS 2005 Matthias Rüther 23
Lens Iris
Depth of field limits:– Wd = working distance– Bs = size of blur spot– I = amount of iris aperture– F = focal length
2**1ffwd
Ibs
wdDOF
e.g.: a 10mm wide object is imaged on a 1/3“ Megapixel CCD from a distance of 100mm, the blurred spot size is max. 5μm
-> best f is 26.5mm, choose 25mm standard lens
-> DOF=0.08mm at full aperture
-> DOF= 0.24mm at aperture = 4
Robot Vision SS 2005 Matthias Rüther 24
Lens types
Standard lenses: focal length from 5mm to 75mm– Adjustable/fixed focus
– Adjustable/fixed Iris
– Adjustable/fixed zoom (focal length)
Macro lenses– Near field imaging (wd ~75mm-90mm, dof ±0.06mm… ±5mm,
magnification 0.14…8)
Telecentric lenses– Parallel projection, moving object towards lens does not change the
image
Robot Vision SS 2005 Matthias Rüther 25
Lighting
Illumination is the most critical part in a machine vision system.
Small illumination changes may severely affect performance of vision algorithms.
If possible, adjust lighting conditions and keep them fixed!
Properties:– Intensity
– Spectrum
– Frequency (amplitude change: flicker, strobe)
– Direction
Hazards:– Object: reflection, specularity, color, stray light, transparency, motion
– Lamp: heat, flicker, stability, lifetime, size, power, speed
Robot Vision SS 2005 Matthias Rüther 26
Regulated Halogen Lamp Systems
Illumination by Quartz-Halogen lamps
High power output
Power control by Voltage regulation and adjustable shutter
Fiber optic light guidance to avoid heating
High power consumption (150W lamp)
Heavy DC power source necessary to avoid flicker
Lamp life: 200-10000hrs
Robot Vision SS 2005 Matthias Rüther 27
Light Emitting Diodes
Possible to generate all primary colors
Bright White LED‘s possible (up to 5W per piece) -> Cooling
Life time: 100000+ hrs
Low power consumption -> Small DC current source
Small/light housing
Fast strobe (time limited by driver circuit, down to 1μs pulses)
Packed in LED arrays
Robot Vision SS 2005 Matthias Rüther 28
Types of Illumination
Directional
Glancing
Diffuse
Robot Vision SS 2005 Matthias Rüther 29
Types of Illumination
Ring Light
Diffuse Axial
Brightfield/Backlight
Robot Vision SS 2005 Matthias Rüther 30
Types of Illumination
Darkfield
Structured Light (Line Generators)