8
9/10/2013 1 Won Suk “Daniel” Lee Professor Agricultural and Biological Engineering University of Florida 1 Nondestructive sensing technologies Nearinfrared spectroscopy (NIRS) Timeresolved reflectance spectroscopy (TRS) Machine vision Multispectral / hyperspectral imaging Magnetic resonance imaging (MRI) Fluoroscopy with Xray Applications NIRS applications Disease detection: citrus greening / citrus black spot Xray & MRI: fruit seed count 2 3 Reflected or Emitted Transmitted Absorbed Incoming energy 4 Spectrometry measures information related to the chemical composition of the materials with the wavelength 5 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 0 10 20 30 40 50 60 70 80 90 Wavelength (nm) Reflectance (%) Tangelo Hamlin Valencia Average fruit spectral reflectance over a growing season from July to January (Kane & Lee, 2006) The concentration of an absorber is directly proportional to the sample absorbance 6

Lee-Non-destructive technology-v2 - University of Florida

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

9/10/2013

1

Won Suk “Daniel” Lee

Professor

Agricultural and Biological Engineering

University of Florida

1

Non‐destructive sensing technologies Near‐infrared spectroscopy (NIRS) Time‐resolved reflectance spectroscopy (TRS) Machine vision Multispectral / hyperspectral imaging Magnetic resonance imaging (MRI) Fluoroscopy with X‐ray

Applications NIRS applications  Disease detection: citrus greening / citrus black spot X‐ray & MRI: fruit seed count

2

3

Reflected or

Emitted

TransmittedAbsorbed

Incoming energy

4

Spectrometry measures information related to  the chemical composition of the materials with the wavelength  

5400 600 800 1000 1200 1400 1600 1800 2000 2200 24000

10

20

30

40

50

60

70

80

90

Wavelength (nm)

Ref

lect

ance

(%

)

Tangelo

HamlinValencia

Average fruit spectral reflectance over a 

growing season from July to January

(Kane & Lee, 2006)

The concentration of an absorber is directly proportional to the sample absorbance

6

9/10/2013

2

400 500 600 700 800 900 1000 1100 1200 13000

10

20

30

40

50

60

70

80

90

Wavelength (nm)

Ref

lect

ance

(%

)

12 July 2005

20 Sept 20052 Nov 2005

13 Dec 2005

(Kane & Lee, 2006)

7

Stepwise multiple linear regression

Principal component regression (PCR)

Partial least squares (PLS) regression

Hierarchical dimension reduction 

Kullback‐Leibler divergence (KLD)

Develop prediction models (calibration)

8

Applications: 

Crop properties (nutrients, water, disease)

Soil properties (organic carbon, moisture, mineral nitrogen, phosphates, pH, particle size)

Weed detection9

A widely used tool for detecting soil and crop properties utilizing spectral signatures

Advantages: non‐destructive measurement and easy sample preparation

10

White & red grapefruits from Florida, ‘Honey’ tangerines

11

Valencia & Hamlin oranges, Thompson red  grapefruits 

12

9/10/2013

3

Healthy  Healthy  Symptomatic Symptomatic

13

0

10

20

30

40

50

60

70

80

90

200 600 1000 1400 1800 2200 2600

Reflenctance (%)

Wavelength (nm)

Healthy 1

Healthy 2

Black spot 1

Black spot 2

Black spot 3

Black spot 4

Black spot 5

Black spot 6

Black spot 7

Avg of Healthy

Avg of Black spot

14

Most commonly used method for quality evaluation

Cost effective

Mature technology

Still needs more improvement for actual applications

15 16

(Valero et al. 2004)

17

Using the light diffusion theory, light absorption and scattering coefficients are obtained. 

Simultaneous determination of chemical concentration & firmness

More novel than standard spectroscopy

Need a photon counting sensor (~$25,000)

18

9/10/2013

4

(Valero et al. 2004)

19 20

21 22

23 24

9/10/2013

5

25

32 cherries / sec

http://www.gpgraders.com.au/products/cherry‐grading/

30 images / cherry

26

OriginalHue

Hue thresholded and inverted

Fruit only

27

(adapted from 

Dewdney, 2012)

Final result Original

Sat. thresholded and invertedHue thresholded and inverted

‘AND’

=

28

29

Reflectance for various objects in a citrus grove

0

10

20

30

40

50

60

70

80

90

100

400 900 1400 1900 2400

Wavelength (Nm)

Re

flect

an

ce (

%)

Big leafCitrusSmall leaf

815 nm

1190 nm

+ = ?

30

9/10/2013

6

(Source: http://www.sciencedirect.com/science/article/pii/S003442570900073X)

31

Common applications – yield mapping, nutrient & disease detection, water stress, …

Immature green citrus fruit detection

Citrus greening disease (HLB) detection

Blueberry fruit detection

32

Actual fruit locations

Pixel segmented image

Detected fruitHyperspectralimage

RGB image

Okamoto, H., and W. S. Lee. 2009. Green citrus detection using hyperspectral imaging. Computers and Electronics in Agriculture 66(2): 201‐208.

33

(Source: R. Ehsani)

34

Hyperspectral imaging combines  images with spectral information.

Multispectral imaging is more suitable for practical applications due to cost and processing time.

35 36

9/10/2013

7

How MRI works? (2.5 min.)(http://www.youtube.com/watch?v=pGcZvSG805Y) 

(Source: http://insideinsides.blogspot.com/2010/07/orange.html)

Orange

37

(Source: http://insideinsides.blogspot.com/2010/12/persimmon.html)

38

Non‐invasive & non‐destructive method

Not suitable for conveyor systems

Takes long time for imaging

Need much maintenance

Safety Strong magnetic field ‐ avoid any metal objects!

Heating caused by radiowave absorption

Cost Conventional system: $1 – 3 million

Portable unit (Prepolarized MRI): $50,000 (http://www‐

mrsrl.stanford.edu/research.html#top)

39 40

X‐ray imaging technique for acquiring real‐time moving images

Low energy is sufficient for agricultural crops

Instrument: less than $100,000

41

(Source: http://spie.org/x84974.xml, 2012)

Seed

“Not possible to discriminate between seeds and pulp consistently using CT or x‐ray images because of their similar densities…”

42

9/10/2013

8

(Source: http://spie.org/x84974.xml, 2012)

• 0.2 T used• Took 7 sec

43

Many non‐destructive sensing technologies are available

Each has its own advantages and limitations

NIRS, multispectral imaging, and X‐ray have great potential and promising… 

44

45