2. PR O B LEM STA TEMENT
Previous studies have
shown that fertilization is usually the largest component of
oil-palm production costs.
According to Goh and Hardter ,
fertilizers in oil-palm plantations can represent between
50%
and 70% of field operational costs, and about 25% of the
total production cost.
● Therefore, a strategic objective of crop
management is to apply nutrient management early on.
● Nowadays, nutrient deficiencies are identified
by trained personnel through foliage analysis and visual
in_x0002_spection in oil-palm plantations.
● However, these methods are
expensive and they are prone to error
3. CR O P NU TR ITIO N IN F ER TILIZ ER
R ED U CTIO N
Crop nutrition is an essential task to obtain high yields and
high-quality products in intensive agricultural strategies ,
which makes fertilizers a requirement to guarantee fertile soils.
However, inadequate usage of fertilizers causes water and soil
pollution, and jeopardizes human health and wildlife. Modern
nutrient management guidelines advocate for a reduced application
of inorganic fertilizers but this requires an estimation of
optimal fertilizer rates that considers the spatial and temporal
variability of crops.
4. MINER A LS TH A T A R E SCA R CE
SY MPTO MS
PO TA SSIU M
NITR O G EN CHLOROSIS
MA G NESIU M
ORANGE SPOTS,BLOTCHES,WHITE STRIPES
ORANGE FRONDS
5. W H A T?
PalmHAND, an application
for real-time identi-
fication of nutrient
deficiencies in oil-palm
crops that can
contribute to control
fertilizers.
This application can be
scaled
to handle multiple users in
oil-palm plantations
spread out
in vast areas, allowing
producers to increase
economic
revenue and reduce
resource usage.
PalmHAND uses visual
symptoms in the oil-palm
foliage to
identify nutrient
deficiencies through
digital image processing
and pattern recognition
algorithms within a smart
device
PA LMH A ND ? ?
6. STEPS TO SO LU TIO N
digital image processing is a mature field
that has been applied successfully to the
identification of plant
diseases
a mobile
application allows the usage of a widely spread
and proven
smartphone technology rather than other
customized and often
expensive equipment.
this application allows realtime monitoring and
assessment of large-scale plantations
because nutrient-deficiency predictions of oil palms are
geo_x0002_located, uploaded to Microsoft Azure, and
displayed using
Power BI reports
01 02 03
7. STEP 1 :
The first step to train the classifier using supervised learning
involved acquiring and labeling sample images of palms
in four categories of interest: healthy, and with deficiencies of
K, Mg, and N respectively. The Colombian Oil Palm Research
Center (Cenipalma) kindly provided 52 labeled images of
oil palm leaves that were shuffled and divided into three
non-overlapping datasets corresponding to: training (∼ 60%),
model selection (∼ 10%), and test (∼ 20%)
8. STEP 2:
The resulting model was stored in a file that
was later transferred to the smart device and
read by the mobile application.
9. STEP 3:
Each observation that is sent to
the cloud contains: a unique device id, a timestamp (year,
month, day, hour, minute, second), a position stamp (latitude,
longitude, and altitude), and four variables with decimal values
between 0 and 1 inclusive. These values correspond to the
probabilities predicted by the classifier, indicating the degree
of confidence that the oil palm captured in the image has a
deficit of Mg, N, K, or that it is healthy.
10. The visualization component was developed
using Microsoft Power BI. This acts as the interface for the
Azure SQL database, transforming the stored data into an
interactive report that allows the user to get real-time insights
about the nutrient deficiencies of the oil-palm plantation being
inspected.
11.
12. Conclusion
This paper introduced PalmHAND, a novel application for
smart devices that allows farmers to get in-situ identification
of nutrient deficiencies of Mg, K, or N, in oil-palm plantations
using leaf photos. If an Internet connection is available, the
application acts as an IoT device in Microsoft’s Azure Cloud
to get storage and visualization of historic data that can
be collected by multiple users working at the same time.
PalmHand was conceived and developed to aid fertilization
management in the oil-palm plantations in remote areas, where
trained experts are not always available
13. TEA M MA TES:
LA V A NY A M
G U R U V A R SH INI B