Machine Vision

1
Machine Vision
(Black and White)

• Autumn Glenn
• 10/15/08

2
Definition

• Machine vision is the capturing of an image (a
  snapshot in time), the conversion of the image to
  digital information, and the application of
  processing algorithms to extract useful
  information about the image for the purposes of
  pattern recognition, part inspection, or part
  positioning and orientation Ed Red

3
Definition

• Black and White (Binary) Machine Vision
• Light intensity of each pixel is reduced to
  either black or white
• Grayscale system
• Usual use 4, 6, or 8 bits of memory
• 8 bits 28 256 intensity levels

4
Current State

• 1970s - first use of image processing
• 1980s grayscale algorithms developed,
  industrial applications cameras
• 1990s massive growth, smart cameras available,
  PC technology used
• 2000s FireWire digital camera technology is
  adopted, market expands rapidly
• Now Used in almost every industry, beginning to
  use color, getting smaller and more compact,
  smart cameras

5
Smart Camera

• Self-contained vision system
• Contains
• Image sensor
• Communication interface
• Image memory
• Processor
• RAM
• I/O lines
• Lens

6
Uses

• Who
• Agriculture, Research, Medical, Automotive
• What
• Inspection, Comparison
• Where
• Production lines, Packaging, Hospitals
• When
• Intricate, Tedious, Dangerous, Uncomfortable

7
Requirements

• Digital Camera(s) camera interface and
  processor
• Input/output hardware
• Frame grabber
• Lenses
• Light sources
• Trigger

8
Cost

• Smart camera 2,000-4,000

www.ni.com
9
Rules and Limits

• Segmentation
• Define and separate regions of interest
• Thresholding
• Convert each pixel into binary (B or W) value by
  comparing bit intensities
• Edge detection
• Locate boundaries between objects
• Feature extraction
• Determine features based on area and boundary
  characteristics of image
• Pattern recognition
• Identify objects in midst of other objects by
  comparing to predefined models or standard values
  (of area, etc.)
• Lighting
• Front
• Highlighting surface items
• Back
• Contrast

10
Primary Vendors
11
Standards
European Machine Vision Association EMVA
Standard Compliant 1288 an initiative to define
a unified method to measure, compute and present
specification parameters for cameras and image
sensors used for machine vision applications.
12
Applications

• Inspection
• Measurements
• Verification
• Detection of Flaws
• Seals
• Guidance and Control
• Autonomous Vehicles
• Identification
• Face Recognition

13
Technical Paper

• Automatic Inspection System Using Machine Vision
• Applications of machine vision for inspection
• Threshold levels
• Optimize time by varying calculation steps and
  methods

14
Video

• http//video.google.com/videosearch?ieUTF-8oeUT
  F-8sourceidnavclientgfns1qmachine20vision2
  0inspectionum1saNtabwv

15
Example

• The Pixel count of a solid-state camera is 500 X
  582. Each pixel is converted from an analog
  voltage signal to the corresponding digital
  signal by an analog-to-digital converter. The
  conversion process takes .08 microseconds to
  complete. Given this time, how long will it take
  to collect and convert the image data for one
  frame? Can this be done 30 times per second?

16
Example
The Pixel count of a solid-state camera is 500 X
582. Each pixel is converted from an analog
voltage signal to the corresponding digital
signal by an analog-to-digital converter. The
conversion process takes .08 microseconds to
complete. Given this time, how long will it take
to collect and convert the image data for one
frame? Can this be done 30 times per second?

• 500 X 582 291,000 pixels
• Tconvert291,000 pixels .08 pixels/s 10-6
  .02328 s
• 1/.02328 s 42 times

17
Summary

• Machine vision is very versatile
• It can be and is used in almost every industry
• It is evolving quickly

18
References

• http//www.vmv.com.au/
• Howison, Robert, When is Colour Required by
  Machine Vision?, DALSA, 2007
• http//www.machinevision.co.uk/