Image Acquisition. Jos J.M. Groote Schaarsberg Center for Image Processing

Transcription

1 Image Acquisition Jos J.M. Groote Schaarsberg

2 Specification and system definition Acquisition systems (camera s) Illumination Theoretical case : noise Additional discussion and questions CBP course: Image Acquisition 2

3 Specification en system definition (1) Which property do you want to know? Existence, diameter, thickness, colour, shape? Or only a deviating parameter e.g. absence or too small radius, too thick, deviation of hue, or misalignment? Do you want a measurement of a property, classification, or only reject bad products? What are the kinds of products? Man-made metal objects, or maybe potatoes? What are the ranges of size and shape? Which are the properties you can use to distinguish the feature of interest? CBP course: Image Acquisition 3

4 Specification en system definition (2) What about the handling of the objects Is it possible to align them neatly in your laboratories Or do they pass on a conveyor belt in all kind of orientation and mixed with other products? What are the environmental conditions? Probably you have to expect straylight, mechanical vibrations, a lot of dust. But not on Mondays: the conveyor belt has been cleaned and is causing some disturbing shiny reflections! CBP course: Image Acquisition 4

5 Specification en system definition (3) High speed conveyor belt for mass production Is the acquisition speed of a common CCD camera sufficient? Or is it just sufficient if it takes one image of two subsequent objects at the same time! Is the processing speed of the computer sufficient, or is it just a much to complex algorithm? CBP course: Image Acquisition 5

6 Specification en system definition (4) What is the use of the acquisition system and its output?? Are we talking about an unique laboratories equipment with a multi-purpose function? Are we talking about 10 dedicated systems in order to automate an assembly line? Or even 200 systems along the highways for the tax administration? CBP course: Image Acquisition 6

7 Use this kind of questions Determine the physical properties which have to be measured Select the best type of sensor Select an appropriate illumination Select the mechanical handling Hardware image acquisition and (pre-) processing Describe the software processing Describe system control including User interface Communication Signalling en report facilities CBP course: Image Acquisition 7

8 Image acquisition systems (camera s) Type of sensors Si CCD matrix camera (video or still image camera) Rectangular, recognisable image Geometrical very stable Cheap Many models CBP course: Image Acquisition 8

9 Image acquisition systems (camera s) relative response Silicium detector relative spectral response eye Si-CCD Si with KG wavelength(nm) CBP course: Image Acquisition 9

10 Image acquisition systems (camera s) Alternative systems (With their advantages and disadvantages with respect to the CCD-matrix): Si CCD line-scan camera (including scanners) Difficult to recognise image Very high resolutions Very suitable for moving objects Photo-diode array Less noise High dynamic range Often large pixels CBP course: Image Acquisition 10

11 Image acquisition systems (camera s) (Near) infrared camera Different spectral response Expensive 3D sensor e.g. triangulation Specific for this one (1) property CBP course: Image Acquisition 11

12 Glossary (1) Integration period 1/50 sec Electronic shutter Saturation (pixel) storage Output amplifier Blooming and smear Dark signal Thermally induced DSNU / crystal defects CBP course: Image Acquisition 12

13 Glossary (2) PRNU: photo response non-uniformity Local variations in responsitivity Often due to layer thickness Mostly wavelength dependent For practical purposes: the same amount of light causes a slightly different signal Gamma Spectral response Number of lines CBP course: Image Acquisition 13

14 Finding your way (1) CBP course: Image Acquisition 14

15 Finding your way (2) CBP course: Image Acquisition 15

16 Finding your way (3) CBP course: Image Acquisition 16

17 Illumination Type of illumination Diffuse lightsource Directed lightsource Specular and/or diffuse reflection Surface illumination Transmission image Backlighting/shadow image CBP course: Image Acquisition 17

18 Lightsources (1) Halogen With reflector Cold light With silver reflector Many variant Small bulb and small coil Broad spectrum Fluorescent strip light Large Cheap Diffuse lightsource CBP course: Image Acquisition 18

19 Lightsources (2) Xenon Very bright Very small spot Also output in the UV spectrum Laser Directed lightbeam Applicable with interferometry CBP course: Image Acquisition 19

20 Lightsources (3) 150 xenon Lightsources source A relative output TL white TL daylight source C wavelength (nm) CBP course: Image Acquisition 20

21 Lightsources (4) CBP course: Image Acquisition 21

22 NOISE Introduction Example of noise due to acquisition A few noise-sources of CCD-chip and camera Basic techniques Demonstration: When the image acquisition has been excellent, the data processing can be much more simple and reliable CBP course: Image Acquisition 22

23 Ideal image Greyvalue pattern with increments of 5%. CBP course: Image Acquisition 23

24 CCD-noise Artificially some noise has been introduced, using typical values of a reliable, commerciallyavailable CCD-chip CBP course: Image Acquisition 24

25 Noise-reduction Based on a-priori knowledge, the noise has been reduced using a uniform filter in the direction which the greyvalue has to be steady. CBP course: Image Acquisition 25

26 Ideal image Greyvalue pattern with increments of 5%. CBP course: Image Acquisition 26

27 CCD-noise Histogram of some artificial noise which has been introduced, using typical values of a reliable, commerciallyavailable CCD-chip CBP course: Image Acquisition 27

28 Noise-reduction Based on a-priori knowledge, the noise has been reduced using a uniform filter in the direction which the greyvalue has to be steady. CBP course: Image Acquisition 28

29 Sources of "noise and other irregularities (1) Natural variations of objects Isolate characteristic properties Take many test samples Lightsource/optics Consult specialist Record and correct White reference / black reference Environmental conditions Minimise Filters/mask CBP course: Image Acquisition 29

30 Sources of "noise and other irregularities (2) CCD-chip LSB (1-sigma 1/3-2/3 DN) Photo Response Non-Uniformity (3-sigma 12%) Shot-noise (1/ (e-) at saturation 2e5 e - ) Signal-independent e.g. dark_current, read-noise and reset-noise (5-10 mv at typical 1.2 mv/dn) Blooming/Smear CBP course: Image Acquisition 30

31 Sources of "noise and other irregularities (3) Camera electronics MTF (pixelclock 15 MHz; bandwidth CCIR 5MHz) Non-linear response Signal transport/recording (see Camera electronics and A/D conversion) A/D conversion Non-linear response (AGC; γ) Phase-Jitter ( Y/N pixelclock) CBP course: Image Acquisition 31

32 Basic techniques (1) Isolate and optimise characteristic properties Be suspicious with respect to natural variations of the object. Minimise environmental influence Do not trust automatic features e.g. AGC, γ,... Use a so-called white-reference and blackreference to reduce fixed-pattern noise. Use average values from multiple images to reduce temporal noise CBP course: Image Acquisition 32

33 Basic techniques (2) Produce robust data-processing i.e. do not depend on even 1% difference between the good, the bad and the ugly. (A-priori) knowledge! Ask questions! Increase knowledge! Use it! CBP course: Image Acquisition 33

34 Ideal image CBP course: Image Acquisition 34

35 CCD-noise CBP course: Image Acquisition 35

36 References: Books Batchelor, B.G. e.a.; Automated Visual Inspection; ISBN Holst, Gerard C.;CCD Arrays, Cameras and Displays; ISBN Hobbs, Marvin; Video Cameras and Camcorders; ISBN CBP course: Image Acquisition 36

37 References: some URL s CBP course: Image Acquisition 37