New Algorithms to Improve X-Ray Inspection

Principles*of*x.ray*inspection

2

The*first*100*years*in*x.ray*history*focused*on*hardware*improvements

3

This*century,*algorithms*are*driving*x.ray*innovation

• Significant*improvement*in*computing*power

• Explosive*algorithm*focus*– see*Google• Marginal*improvement*in*hardware

4

Outline

• SMT*Defects*identified*with*x.ray*inspection

• WISE• Dual*Energy• ICARUS• Artificial*Neural*Networks

5

Algorithms*for*X.Ray*Inspection

6

RAW IMAGE REAL TIME ALGORITHMS

REAL TIME IMAGE

WISE:*Wavelet*Image*Spectra*Enhancement*

• X.ray*inspection*enhancement*algorithm*based*on*parametric*estimation*of*image

• Runs*in*real*time*– not*an*offline*filter• What*is*parameter*estimation?

PATENT PENDING

Raw datas = A.sin(2πf + ϕ)A = 2f = 4kHzϕ = π

Parameter Estimation

8

From*Wavelet*transform*to*Parameter*Estimation

Wavelet*Transform

Parameter*Estimation*Model

9

Results*before*and*after*WISE

10

Surplus*stock*of*electronic*components*dumped*into*the*market*further*establishing*the*broker*and*grey*marketsHuge*amounts*of*scrap*electronicsComponents*made*obsolete*Need*for*short*delivery*timeMassive*cost*reduction*pressures*on*OEM*buyers*for*double*digit*price*reductionsBuying*and*selling*via*the*internet*became*common*practice

The*counterfeit*electronics*problem

11

Component*Harvesting

12

US*Dept*of*Commerce*Report

13

Only*expensive*parts*are*counterfeited…

Anything*wrong*with*these*caps?

15

What*happens*when*you*got*the*wrong*cap?

16

17

18

Dual*Energy*X.Ray*Imaging

PATENT PENDING

Low Energy Image High Energy Image

19

Dual*Energy*X.Ray*Imaging

PATENT PENDING

Dual Energy Image

20

BGA*with*voids*– dual*energy

21

PATENT PENDING

Transformer*– dual*energy

22

How*can*we*use*these*innovations*for*counterfeit*detection?*What’s*wrong*

here?

23

How*can*we*use*these*innovations*for*counterfeit*detection?*And*here?

24

How*can*we*use*these*innovations*for*counterfeit*detection?

25

ICARUSImage Comparison Algorithms for Radiography Unified Software

26

ICARUS*Comparison*Algorithm

• Challenge*is*to*find*the*global*minimum*

27

Unfortunately*things*are*not*simple…**

• Multiple*algorithms*compete*for*best*answer

28

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

1st – SELECT REFERENCE

IMAGE

2nd – DEFINE COMPARISON

MASK

29

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

4th – CORRECT FOR ROTATION AND OFFSET

TRANSLATION

3rd – COLLECT IMAGE FROM

TAPE

30

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

5th – MEASURE SIMILARITY BETWEEN REFERENCE AND ACQUIRED IMAGE WITH

MASK APPLIED TO BOTH IMAGES

REFERENCE ACQUIRED

31

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

6th – COMPARE MEASURED SIMILARITY WITH SET THRESHOLD

REFERENCE ACQUIRED

Example: 98% SIMILARITY: IF < THRESHOLD ! FAILIF > THRESHOLD ! PASS

DIFFERENCES

32

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

REFERENCE ACQUIRED + DIFFERENCES

33

ICARUSImage*Comparison*Algorithms*for*Radiography*Unified*Software

7th– REPEAT PROCESS FOR WHOLE TAPE. STOP ON FAIL

34

Conclusions

• Innovation*in*the*first*century*of*the*x.ray*inspection*market*was*driven*by*hardware*improvements

• Massive*computing*power*has*allowed*the*second*century*to*be*driven*by*algorithm*innovations

• Expect*to*see*more*algorithm*improvements*this*century

35