Professional Documents
Culture Documents
SOIL SENSORS
Soil Sensors: the Need of the Hour
• Can transform global agriculture
• Right source of plant nutrient, in right rate, right time and right place
Digital
agriculture
• A decade of research in
• Rapid natural resource management via generation of spectral libraries of soil, crop and water
• Development of microfluidics, MEMS and new-age sensors
• Digital mapping and geospatial modelling
Preparedness
PXRF: Advanced Portable Sensor
• Energy released by replacing of
inner shell electrons by outer shell
electrons (termed fluorescence) is
quantified by a silicon drift detector
in the aperture of the instrument .
• Scanning (dwell) time for a given
sample is typically ∼60–90 s.
• Advantages:
• Field portability
• No consumables
• Non-destructive
• Multi-elemental analysis
(providing simultaneous analysis
of ~20-30 elements)
• Parameters:
• Soil pH, EC, CEC, P, K, Ca, Mg, S,
Micronutrients, Gypsum, %BS,
Heavy metals, permafrost pH,
LULC, Parent material, Profile
horizonation, Geochemistry,
Compost EC, Compost CEC, Water
heavy metals, Leaf elements
Reference Methods
Soil Science
USDA NRCS: 7.3.1 X- Society of
US-EPA Method Ray Fluorescence, America
6200 Handheld
XRF Spectra
2 New approaches
1.Combining XRF beams as such
2.Selecting the peak and extracting the FWHM and Hmax
PXRF for LULC
Forests (147 samples), agricultural
lands (149 samples), and iii) converted
lands (converted from forest to
agriculture over the past 20 years) (127
samples).
Classification Results
CART
Other PXRF Work Published
SEM-EDAX
Diffuse Reflectance Spectroscopy
• Fast
• Cost-effective (almost zero recurring cost)
• Needs minimum or no sample pretreatments
• High Throughput and Non-invasive
• Portable
• Parameters:
• OC, Available N, P, K
• pH and clay
• Moisture and heavy metals
VisNIR DRS
Soil Material
VisNIR-DRS Equipment
Scissoring
Asymmetrical stretching Symmetrical stretching
Holographic grating
Fiber
1.4-2 nm
Separation filter
512 Si photodiode array
Light
InGaAs detector
~600 channels
InGaAs detector
~600 channels
2 nm Source/probe
Reflection Light
Sample
Converts incident photons to
Digital Voltage electrons (photocurrent)
signal
16 bit A/D converter
Working Mechanism
Spectralon Panel
❑ 99% reflectance
❑ Optimize spectrometer
❑ White reference: simplifies reflectance
❑Removes ambient temperature effect
❑ Removes humidity effect
❑ Optimize : probes change
www.asdi.com
Spectral Resolution
Spectral Preprocessing
Raw scan Seamless spectra
❑ Increasing resolution
❑ Background elimination
❑ Noise filtration
http://www.youngin.com/application/AN-0608-0115EN.pdf
Why Derivative Spectra
http://www.youngin.com/application/AN-0608-0115EN.pdf
t t
Fourier Transform
Fourier Series (cont’d)
α1
α2
α3
Fourier Transform
Continuous Fourier Transform (FT)
where
Fourier Transform
❑ SEE VIDEO
https://www.youtube.com/watch?v=spUNpyF58BY
Wavelet Transformation
Motivation
time, t
Wavelet Analysis is based on an short duration
wavelet of a specific center frequency
time, t
Wavelet
A wavelet is a mathematical function
useful in digital signal processing
and image compression
Addison, 2002
Types of Wavelet
Wavelet Transformation
❑ Wavelet transform of CWT
1 +∞ 𝑡−𝑏
T(a,b)=√𝑎 −∞ 𝑥(𝑡)ψ∗ ( 𝑎
)dt
❑ Discrete wavelet transform
1 𝑡−𝑛𝑏𝑜𝑎𝑜𝑚
ψm,n (t)= ψ( 𝑎 𝑚 )dt
√𝑎𝑜𝑚 0
Addison, 2002
Wavelet Transform
coefficients
of wavelets
coefficient of wavelet
complex conjugate of
with
wavelet with
scale, s and time, t
scale, s and time, t
Wavelet
normalization
shift in time
change in scale:
big s means long
wavelength
wavelet with
scale, s and time, t
Mother wavelet
Shannon Wavelet
mother wavelet
t=5, s=2
time
Fourier spectrum of Shannon Wavelet
frequency, w
= f(t) * Y(-t/s)
sj=2j
Scale changes
and
by factors of 2
tj,k = 2jkDt
where j = 1, 2, …
k = -… -2, -1, 0, 1, 2, …
dyadic grid
The factor of two scaling means that the spectra
of the wavelets divide up the frequency scale into
octaves (frequency doubling intervals)
w
1/
8wny ¼wny ½wny wny
As we showed previously, the coefficients of Y1 is just the
band-passes filtered time-series, where Y1 is the wavelet, now
viewed as a bandpass filter.
This suggests a recursion. Replace:
w
1/
8wny ¼wny ½wny wny
with
low-pass filter w
½wny wny
And then repeat the processes, recursively …
Chosing the low-pass filter
It turns out that its easy to pick the low-pass filter, flp(w). It must
match wavelet filter, Y(w). A reasonable requirement is:
|flp(w)|2 + |Y(w)|2 = 1
That is, the spectra of the two filters add up to unity. A pair of such
filters are called Quadature Mirror Filters. They are known to have
filter coefficients that satisfy the relationship:
tj,k = 2jkDt
we merely subsample the time-series by a factor of
two after each filtering operation
time-series of length N
HP LP
Recursion for
wavelet
2 2
coefficients
g(s1,t)
HP LP g(s1,t): N/2 coefficients
2 2
g(s3,t) …
Wavelength ( N) Filter Bank
HP LP g(s11,t): N/2 coefficients
g(s10,t): N/4 coefficients
2 2 g(s9,t): N/8 coefficients
Total: N coefficients
g(s11,t) Wavelet coefficients at scale 7,6,5 were
HP LP extracted
Stepwise Multiple linear regression
2 2
g(s10,t)
HP LP
2 2
g(s9,t) …
Coiflet low pass filter
time, t
Coiflet high-pass filter
time, t
From http://en.wikipedia.org/wiki/Coiflet
Spectrum of low pass filter
frequency, w
Spectrum of wavelet
frequency, w
time-series
stage 1 - hi
stage 1 - lo
Stage 1 lo
stage 2 - hi
stage 2 - lo
Stage 2 lo
stage 3 - hi
stage 3 - lo
Stage 3 lo
stage 4 - hi
stage 4 - lo
Stage 4 lo
stage 5 - hi
stage 6 - lo
Stage 4 lo
stage 5 - hi
stage 6 - lo
Had enough?
Putting it all together …
|g(sj,t)|2
short
wavelengths
scale
long
wavelengths
time, t
Wavelet Thresolding
Noise corrupted spectra Haar wavelet transform
Spectral
preprocessing
database select make model
selected transformed
data
data
Interpretation&
data prospecting
rule formulation
and surveying
SOC SIC
0.4
Reflectance
0.3
PCA Screeplot
Soil As Map
Soil As Pools by DRS
Significant Wavelengths
Spectrum
Portable VisNIR Camera
Upgraded version
Soil OC box + Smartphone
2.5
y = 0.52x + 0.32
2 R² = 0.84 The images acquired were of dimension
Measured OC %
NixPro
USB microscope image+ deep learning for soil OC classification
0.25
B value
1 0
-0.25
PC2
0
-4 -2 0 2 4 -0.5
-1 -0.75
-1
-2 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
F1 (75.72 %)
-3 Active variables
PC1
Conclusions