Colorado School of Mines
Computer Vision
Professor William Hoff
Dept of Electrical Engineering &Computer Science
Colorado School of Mines
Computer Vision
http://inside.mines.edu/~whoff/
1
Other Type of Image Sensors
Colorado School of Mines
Computer Vision
Color Images
• Each pixel is a triplet of red, green, blue values (RGB)
• To save storage space, the images can be stored with a “colormap”, which is a lookup table
– The image values are the indices into the colormap
– Colormap specifies RGB for each index
Colorado School of Mines
Computer Vision
3
Color Sensors
• The CCD sensor array has colored filters over each sensor element
from “Computer Vision:
Algorithms and Applications” by Richard Szeliski
• RGB values are interpolated so that we have values at each pixel
Colorado School of Mines
Computer Vision
4
Range Sensors
• Each pixel value is the range to the nearest surface in the scene
• Range images are also called depth maps, or 2.5D images
• These can be combined with RGB cameras; to create color+depth images (called RGBD cameras)
• Types
– Structured light systems
– Time of flight ‐ LIDAR (Light detection and ranging)
Colorado School of Mines
Computer Vision
5
Structured Light (Active Triangulation)
P(X,Y,Z)
Camera
Laser Line Projector
Conveyor Belt

Laser Lines
b
f
Two equations, two unknowns (X,Z)
(1)
(2)
x
x/f = X/Z
tan  = Z/(b+X) or cot  = (b+X)/Z
X
x
b
 
 
Y

  f cot   x  y 
Z
f
 
 
Solve for X,Z:
X = Z x/f
cot  = (b+Zx/f)/Z, Z cot  = b + Zx/f
Z(cot ‐ x/f) = b, Z = b/(cot ‐ x/f) Colorado School of Mines
Computer Vision
6
Structured light sensor to measure surfaces of drums
Intelligent Mobile Sensing System (IMSS) – Martin Marietta, 1991
Colorado School of Mines
Computer Vision
7
Example – Kinect camera
• An infrared speckle pattern is projected onto the scene
• Depth is estimated using triangulation
Colorado School of Mines
RGB‐D images are produced
Computer Vision
8
Time of Flight Systems
•
Pulse systems
– Directly measure time of flight of a very short pulse – Many pulses are fired sequentially and the average response is used
– Requires very accurate sub‐nanosecond timing circuitry
•
Interferometric systems
– Measure the difference in phase between the sensed reflected beam and a reference beam
– There is an “ambiguity interval”
•
Sensors often produce an amplitude measurement as well as a range measurement
Colorado School of Mines
Computer Vision
9