Image Based Measurement

1
Image Based Measurement

• Heimo Ihalainen

2
Image Based Measurement Definition

• Using images ( 2D-data) and computational
  methods to produce measurement results
• Measurement uncertainty and measurement
  information are important issues as in all
  measurement
• Using simple or demanding methods depending on
  measurement task at hand
• Moving or still images

3
Contents

• Areas Close to Image Based Measurement
• Important Technical Development
• Developing Application Fields
• Applications Development at TUT/MIT
• Conclusion

4
Areas Close to Image Based Measurement
image processing
image based measurement
photogrammetry
image analysis
machine vision

• Strong new areas developing medical imaging,
  tomography, image noise analysis/processing,
  laser lighting etc

5
Areas Close to Image Based Measurement

• Image Processing
• Basic image operations, Advanced operations
  often high speed demand
• Moving and still images used everywhere where
  digital images are used
• Special Application Areas Image compression,
  Television and video, Games, Presentation
  graphics
• Special Skills Vast store of methods, Fast
  on-line operation

6
Areas Close to Image Based Measurement

• Image Analysis
• Analysis and feature extraction from images
• Statistical procedures for complex tasks
• High intensity computation
• Mostly still images not limited to camera
  images
• Special Skills New computational methods, stereo
  image analysis, Tomography analysis, Multivariate
  images
• Application Areas Medical image analysis,
  Criminology, Scientific applications Astronomy,
  Application of aerial and satellite images,
  Technical diagnostics

7
Areas Close to Image Based Measurement

• Machine Vision
• Measurements and on-line control based on moving
  image cameras
• Special lighting arrangements often color,
  sometimes laser, lightning effects
• Robot Vision Copying human vision, Artificial
  (machine) intelligence
• Special Skills Fast operation, Intelligent ways
  to arrange light, Stereo vision, Process control,
  Stand-alone application platforms
• Drawbacks Mostly analog cameras, Simple
  processing because of limited time

8
Areas Close to Image Based Measurement

• Photogrammetry
• Spatial measurements based on camera images,
  sometimes with laser lighting and special targets
• Long Range Photogrammetry Measurements based on
  aerial and satellite photographs, Cartography
• Close Range Photogrammetry Industrial 2D- and
  3D-measurements, Measurements at arceological
  excavations, Quality assurance
• Special Skills Stereo and multi camera imaging,
  Image registration, 3D Stereo vision, Measurement
  accuracy analysis, Laser lighting, Camera
  calibration

9
Important Technical Development

• Imaging Technology
• CCD-, CMOS- and layered image sensors
• covering area from infrared to ultraviolet
• matrix- and line-sensors
• color camera technology BAYER-matrix,
  microlenses
• Digital moving image
• fast, low noise sometimes needs special cooling
• still limited by information flow speeds, some
  new buses under development IEEE 1394 (B)
• Digital still cameras
• now market leaders, less noisy than film cameras
• Other electronic imaging
• electronically readable film for electromagnetic-
  and beta-radiation applications
• Computational optics

10
Important Technical Development

• Computers
• Nice speed development
• image based applications are sometimes quite
  demanding for computer speed and memory
• Compact size, Low power usage
• enabling mobile usage
• imaging mobile devices
• Developing 64 bit processors
• memory limit moving from 4 GB

11
Important Technical Development

• Communications
• Internet reaching useful speed for image
  applications
• Computer connections are getting simpler to use
  and faster
• Wireless connections allow mobile operation and
  are useful also in difficult environments

12
Important Technical Development

• Computational Methods (fast development)
• Pattern classification and clustering
• Digital filtering linear and non-linear methods
• Image registration and alignment
• Multivariate image analysis
• Standard computer packages for machine vision
• Tomography, multi image stereology, level set
  methods, transforms, texture analysis
• Computational optics wavefront coding, confocal
  image processing (tomography)

13
Fast Developing Application Fields

• Machine vision
• Computer imaging
• Digital TV and video
• Digital still cameras
• Cameras in mobile phones
• Computational imaging

• Computer Tomography
• Security Applications

14
Developing Application Fields

• Automation of former human tasks in image
  applications
• ex. quality assurance and classification,
  cartography, human activity monitoring
• Totally new application areas
• mobile cameras and GSM/GPRS/3G (digital still
  cameras, video clips) in inspection and
  trouble-shooting

15
Development at TUT/MIT

• Measurement of paper quality
• formation (distribution of basis weight),
• surface form and profile
• printability and print quality
• Flow measurement
• Measurement of wood
• spatial measures
• measurement of quality
• Analysis of camera sensors
• imaging and noise properties

16
Measurement of Paper Quality

• 2D-spectral analysis has been our main method to
  texture analysis

17
Measurement of Paper Quality

• Rings around origo contain variation on different
  scales
• Sectors contain variation in different directions
• 1D-spectrum can be computed by cumulating the
  variation on some axis of 2D-spectrum
• some scale parameters computed from 1D-spectrum
  correlate better with human visual experience

18
Several 2D Measurements

• Example
• board 200 g/m2
• uncoated
• flexo printed
• Measurements (basis weight, pressure load,
  optical scan) made so that they cover partially
  the same area
• Analysis of the variation in formation scale

19
(No Transcript)
20
Analysis

• The dependencies between different measured
  quantities can be displayed
• xy-plot (upper)
• probability density and linearity fit (lower)
• High statistic reliability is achieved because
  there degree of freedom is high

21
Measurement Bubble Size Distribution

• Back light measurement
• Segmentation, size, direction and form analysis
  on a series of images

22
Development of Wood Quality Measurement
23
Using 2D-spectrum

• Cartesian and polar coordinates

24
Thickness field of annual rings
25
Using Annual Ring Texture

• Detecting flaws in texture field

26
Using Images from Side
27
To Detect Quality and Kind

• Textural and color parameters can be used
• This example textural parameters
• horizontal average wavelength
• vertical average wavelength

28
Image Sensor Noise

• Computing photon transfer curve from sample
  images
• On right side Canon D30 (CMOS), Nikon D70 (CCD),
  Sigma SD10 (Foveon)

29
Conclusion

• Image based measurement is in fast development
  phase
• In future large percentage of all measurements
  will be based on images
• non-touching
• good statistical accuracy achievable
• Much development is going on and is still needed
• Connection between sensor technology and image
  applications (TUT/MIT)