Page 1
Optical Inspection in Tool Industry
and Manufacturing
Dr. Daniel Garten1, Dipl.-Phys. Heinz-Wolfgang Lahmann1
and Dr. Katharina Anding2
1) GFE – Society for Production Engineering and
Development
Näherstiller Straße 10, D-98574 Schmalkalden
[email protected]
URL: http://www.gfe-net.de
2) Ilmenau University of Technology
Department for Quality Assurance and
Industrial Image Processing
Gustav-Kirchhoff-Platz 2, D-98693 Ilmenau
[email protected]
URL: http://www.tu-ilmenau.de/qualitaetssicherung
Page 2
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 2 STEINBEISS SpectroNet Collaboration Forum
Outline
1 3-D Inspection of grinding tools
2 2-D Inspection of grinding tools
3 Intelligent methods for 2-D surface inspection
4 Optical detection of production defects at pet-bottles
Page 3
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 3 STEINBEISS SpectroNet Collaboration Forum
Problems of optical inspection in tool industry and manufacturing
- complex objects with a high range of variation
- high variation of surface and optical reflectance characteristics
- rough surrounding conditions (dust, vibrations, foreign light)
- increasing requirements regarding testing time, accuracy and repeatability
- short cycle time
- often 100-%-monitoring required
- . . .
robust image acquisition and image processing algorithms are needed
Page 4
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 4 STEINBEISS SpectroNet Collaboration Forum
Shape-from-Focus with colour information System OMG3
3-D-model
3-D Inspection of grinding tools
Method to measure the highly inhomogenious grinding surface
Page 5
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 5 STEINBEISS SpectroNet Collaboration Forum
3D point cloud for estimating of wear determining parameters
Image acquisition of equally spaced
surface views
Calculation of the contrast value for
every pixel to evaluate the local
sharpness
Find the maxima of the focus function
for every pixel to calculate the height
of the surface
3-D Inspection of grinding tools
Algorithm overview
Page 6
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 6 STEINBEISS SpectroNet Collaboration Forum
3-D Inspection of grinding tools
Grain edges distribution
N
n
SnS zN
K1
*1
galvanic bond
metal carrier
galvanic bond
metal carrier
Grain protrusion
N
n
BnB zN
K1
*1
Parameters to characterize wear on grinding tools
Page 7
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 7 STEINBEISS SpectroNet Collaboration Forum
2-D Inspection of grinding tools
Grinding surface
Aim: Count the abrasive grain and determine its area in the image field of the camera
Found abrasive grain with its calculated area
Gain: 1. Estimate the accuracy of the sieving process,
2. Evaluate homogenity of the spatial grain distribution
Page 8
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 8 STEINBEISS SpectroNet Collaboration Forum
2-D Inspection of grinding tools
Algorithm overview
1. Calibrate the image acquisition system with a micro-structure of known geometry
(in this case a reticle)
2. Choose the red colour channel (different colour models where investigated
regarding its contrast between abrasive grains and bond level)
3. Smooth the histogram of the red channel till only two maxima appear
4. Apply region-growing to detect the connected regions of the grain
5. Fill holes in the detected regions with morphological operations
6. Count the regions and calculate its mean area and area deviation
Peak bond level
Peak abrasive grain
Page 9
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 9 STEINBEISS SpectroNet Collaboration Forum
Intelligent methods for 2-D surface inspection
Detection of surface defects by 3-D-measurement
Example: cone-shaped counterbores aquired with Alicona Infinite Focus SL, after form removal
good surface quality, furrows
only in machining direction bad surface quality, chatter marks
Is it possible to determine surface defects only from
fast and cheap acquirable 2-D texture and color features?
Page 10
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 10 STEINBEISS SpectroNet Collaboration Forum
Intelligent methods for 2-D surface inspection
(currently under development)
Methods for surface inspection based on 2-D texture analysis and machine learning
Page 11
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 11 STEINBEISS SpectroNet Collaboration Forum
Optical detection of production defects at pet-bottles
Aim: Detect water-rings and bubbles at pet-bottles
Idea for solution:
- use a transmitted light setup with an horizontal line structure
- at waterrings or bubbles optical effects cause an distortion of these lines in the image
Page 12
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 12 STEINBEISS SpectroNet Collaboration Forum
Optical detection of production defects at pet-bottles
Algorithm overview
1. Segment the horizontal black lines
2. Segment the region of the bottle
3. Detect the middle line of every segmented region of the black lines with a thinnig
operation
4. Detect endpoints and junctions
5. Count all junctions and endpoints within the region of the bottle, exclude those near
the border, near the bottom and those at the neck (winding)
6. Decision: number junctions/endpoints > 5 => not OK
number junctions/endpoints < 5 => OK
Defect region in detail
Page 13
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 13 STEINBEISS SpectroNet Collaboration Forum
Optical detection of production defects at pet-bottles
Algorithm overview
1. Segment the horizontal black lines
2. Segment the region of the bottle
3. Detect the middle line of every segmented region of the black lines with a thinnig
operation
4. Detect endpoints and junctions
5. Count all junctions and endpoints within the region of the bottle, exclude those near
the border, near the bottom and those at the neck (winding)
6. Decision: number junctions/endpoints > 5 => not OK
number junctions/endpoints < 5 => OK
Defect region in detail
Page 14
26th March 2014 • Dr.-Ing. Daniel Garten • © GFE Schmalkalden e.V. 14 STEINBEISS SpectroNet Collaboration Forum
Thank you for your attention!
Contact:
Dr. Daniel Garten, Dipl.-Phys. Heinz-Wolfgang Lahmann
GFE – Society for Production Engineering and Development
Department for measuring technique and test bench building
Näherstiller Straße 10, D-98574 Schmalkalden
[email protected]
Phone: 03683/690086
URL: http://www.gfe-net.de