AUTOMATIC VISUAL INSPECTION OF ZIRCONITE BLOCKS

jabgoldfishΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 5 μήνες)

50 εμφανίσεις

AUTOMATIC VISUAL INSPECTION

OF ZIRCONITE BLOCKS

raw zirconite blocks

finished zirconites

Z
irconite blocks

: the problem

Raw zirconite blocks are polished and converted into zirconites for use in jewelry.

About 15% of blocks exhibit shape and dimensional defects, which interrupt the

transformation process if they reach the machines.


Human visual inspection is infeasible because
millions

of blocks per month are

processed.


How to view, illuminate and inspect blocks ?

Z
irconite blocks

: approach

I. The first step is the requeriments analysis :
what

must exactly be done,
why

and under


which operational
restrictions

?


Requirements Specification Document.


II. Once we have concluded that . . .



a machine vision approach is feasible


lighting is critical!



machine vision is the best approach



the devised solution may be affordable for the customer and cost effective for


the developer


then



design a system prototype


Preliminary Design Document



evaluate its cost : hardware, software, risk, licenses, copies …



formulate an economic proposal



sometimes prepare a convincing
demo =
implement prototype


III. Propose either . . .



an in
-
depth study (more analysis) that is not for free


Formal Report



to sign a contract and start the project


Contract, Plannification Document, ...

Possible
shape

defects

Z
irconite blocks

: defect characterization

Possible
dimensional
defects



non
-
square base



|
length or width


nominal size
|
>

tolerance,


tolerance = 0.1 mm


nominal sizes from 1.70 to 3.00 mm


(height from 1.25 to 2.00 mm)

Backlight

Z
irconite blocks

: Lighting

FLD from Doal&Jenner (coaxial)

Z
irconite blocks

: Lighting

1

2

3

4

5

6

7

8

9

10

Correct blocks


The “museum of horrors”

Z
irconite blocks

: Lighting

1

2

3

4

5

6

7

8

9

10

11

12

Z
irconite blocks

: Lighting

13

13bis

14

15

16

17

18

19

20

21

22

23

Z
irconite blocks

: Lighting

24

25

26

27

28

Z
irconite blocks

: Lighting

Background : the different of shades of black

velvet

tire rubber

black mate spray paint

We must maximize the difference between backgound and black grey levels so as to

correctly and reliabily segment and measure the blocks.


“Black” must be



really black, not just dark



very diffuse

Z
irconite blocks

: Lighting

2: progressive camera, with


wide open aperture and


shutter 1/1000

5: rectified coaxial lighting

6: fiberoptic photosensor

7: rotation motor and


painted platform

8: manual speed control

3: extensor tube 20mm

4: lens
f
=50mm

1

2

4

1: closed dome with black


painted inner walls

3

5

6

7

8

Z
irconite blocks

: Lighting

fiber optic to photosensor

fiber optic from infrared LED

beamsplitter

Z
irconite blocks

: Lighting

Zirconite blocks

: Defect detection

X
2

= A
1

/ (A
1
+ A
2
)

A
1

A
2

L
1

L
2

D
2

D
1

X
1

= 1


D
2

/ D
1

X
1

= 1


D
2

/ D
1

right

X
1

= 1


D
2

/ D
1

Zirconite blocks

: Defect detection

Z
irconite blocks

What would we do if ...




not only the upper face had to be examined but also
lower or side faces

in order to detect


lack of material not visible from above ?




the
height

of the block must also conform to a certain tolerance ?
Height varies

2 mm.


(depth from focus ?)




6 parts per second

have to be safely inspected ? That means high base speed,


less time to acquire each image or otherwise the camera will yield blurred images




we are later told that blocks can be o several grey shades ?


And how much would the proposed solutions
cost
in terms of hardware and labour time

?

There is not such a thing as a simple machine vision problem :

the risk of a project should never be underestimated