The 2016 MARS Conference will provide a platform to present Member State experiences and other observations regarding the Integrated Administration and Control System (IACS), including developments in shared management.
During the Conference, challenges and opportunities for image use, aspects of performance and the future outlook of the Common Agricultural Policy (CAP) will also be discussed.
The Conference will take place in Lisbon on 24 and 25 November,
Automated classification of land cover for the needs of CAP using Sentinel data. A proactive approach
1. Automated classification of land cover for the
needs of CAP using Sentinel data.
A proactive approach.
PanagiotisILIAS(NP)
p_ilias@neuropublic.gr
DimitrisKAPNIAS(GAIA)
d_kapnias@c-gaia.gr
22nd CAP/IACS conference, 24-25 November 2016, Lisbon - Portugal
2. Farmers’
Organizations
70 Coops
150 K
Farmers
0.75 B €
Turnover
0.5 M ha
Area
Neuropublic
Gaia Cloud
Platform
AgriTech
Know-How
AgriTech
Solutions
Smart
Farming
Piraeus Bank
No 1 GR
Contract
Farming
AgriCard
Young
Farmer loans
The shareholders legacy to the coalition
Farmers
Tech
Bank
3.
4. Society Management
Subsidy
(CAP)
Commerce
Smart
Farming
92 Farmer Service Points (FSP) certified as AID
Application assistants to the beneficiaries
GAIA’s involvement in the Aid Application Scheme
Nationwide
Network of
local service
advisors
(FSP)
GAIA is certified as the coordination &
support body (2014-2020)
6. Atmospheric
Soil Data
GAIA Cloud
RPAS
IoT
GAIA Sense
Satellite Sensors Weather
Data Fusion
Decision Making
Data Analysis
API
GAIA Smart Farm & GAIA Subsidy (CAP)
Advisor
Farmer
Data Collection
Other sources
IoT and Cloud Infrastructure for Agricultural Monitoring
Other Platforms
Other Platforms
large scale agricultural areas
Sentinel-2
• Sort revisit
• HR
• MS
• Open
7. Subsidies control Smart Farming Services
The role of Sentinels
On-the-fly automated cross checks of parcels
declaration (proactive control):
• Reducing errors in the location of parcels
• Discouraging false claims
Full scale automatic compliance cross checks (post-
declaration control):
• crop-wise in terms of modelling and time
windows
• Allows better risk analysis, less RFV, more
effective controls
• Reducing overall error rates
Pest management/Early hazard warnings
Irrigation
Fertilisation
CAP eligibility - Compliance
Sentinel can been used for the
multitemporal extraction of
vegetation, soil and water indices and
the monitoring of the phenological
growth.
Parcel Level
8. Multi-Temporal Object-based Monitoring
Collect
Store
Define
Objects
Assign
Properties
Create
Models
Decision
Making
Take
Actions
Monitor
Actions
[ 0. 0.2 0.5 0.8 0.2 0.7 0.2 0.9 0.7 0. 0.74 0.97
0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.
0. 0. 0.39 1. 0.07 0.01 0. 0. 0. 0. 0. 0. 0.
0. 0. 0. 0.1 1. 0.26 0.06 0.04 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.
0. 1. 0.14 0.03 0. 0. 0. 0. ]
Time
Space
Vegetation – Soil – water Indices at parcel level,
for (single or multi) crop growing period
Crop Models
Parcel = Object
Cloud based services that integrate earth observation data with Image Processing, Machine Learning, Spatial Modeling
9. S2 Collection dates
Extracted Indices patterns agree with phenological growth for the specific
crop (beans)
Multitemporal Visualization of the Prespa area, for the growth period,
using indices values (R=NDVI, G=NDWI, B=SAVI)
Temperature data for area Leukonas - Prespes
Rainfall data for area Leukonas - Prespes
Indices in Agricultural Monitoring
Indices allows us too visualize the behavior of specific crops at parcel level through time.
10. Land Cover Classification using +Indices
b1
b2
b3
bn
Image (b1, b2, b3, … , bn) Object (reference parcel or holding LPIS) Land Cover Class System
Crop Code & Crop Description
1.1.1 xxxxxxxxxxxx1.1.1
Segment
(Polygon: x1y1, x2y2, x3y3, x4y4)
PHASE 2: Descriptor Extraction for each object per Image
Feature Extraction t22
All pixels
All Bands or/ and Indices
SIFT
SURF
ORB
CHIST
Method(s)
[ 0, 0, 0, 32, 0 , 0, …., 123, 0 ]
Descriptor (scaling)
PHASE 1: Indices (i) extraction using atmospherically corrected bands (b)
PHASE 3: Model Creation (data driven)
Train set
Input (time series data for polygons) Output( Labe)l
[ 0, 0, 0, 22, 0, 5, …., 100, 0, 12/04/2016] [1.1.1]
[ 0, 0, 0, 32, 0, 0, …., 123, 0, 07/05/2016] [1.1.1]
…….
[ 4, 1, 1, 22, 0, 0, …., 123, 0, 22/10/2016] [1.1.1]
Training
SVM
RF
NN
Evaluation
PHASE 4: Crop estimation for other parcels using the data-driven input -output model(s)
Input (time series data for a polygon)
[ 0, 0, 0, 22, 0 , 5, …., 100, 0, 12/8/2016 ]
…….
[ 4, 1, 1, 22, 0 , 0, …., 123, 0, 12/8/2016 ]
Land Cover
Classification Model
Model
Multi-Dimensional
Dimension varies
i1
i2
i3
in
Image (i1, i2, i3, … , in)
b1
b2
b3
bn
Image (b1, b2, b3, … , bn) Object (reference parcel or holding LPIS) Land Cover Class System
Crop Code & Crop Description
1.1.1 xxxxxxxxxxxx1.1.1
Segment
(Polygon: x1y1, x2y2, x3y3, x4y4)
PHASE 2: Descriptor Extraction for each object per Image
Feature Extraction t22
All pixels
All Bands or/ and Indices
SIFT
SURF
ORB
CHIST
Method(s)
[ 0, 0, 0, 32, 0 , 0, …., 123, 0 ]
Descriptor (scaling)
PHASE 1: Indices (i) extraction using atmospherically corrected bands (b)
PHASE 3: Model Creation (data driven)
Train set
Input (time series data for polygons) Output( Labe)l
[ 0, 0, 0, 22, 0, 5, …., 100, 0, 12/04/2016] [1.1.1]
[ 0, 0, 0, 32, 0, 0, …., 123, 0, 07/05/2016] [1.1.1]
…….
[ 4, 1, 1, 22, 0, 0, …., 123, 0, 22/10/2016] [1.1.1]
Training
SVM
RF
NN
Evaluation
PHASE 4: Crop estimation for other parcels using the data-driven input -output model(s)
Input (time series data for a polygon)
[ 0, 0, 0, 22, 0 , 5, …., 100, 0, 12/8/2016 ]
…….
[ 4, 1, 1, 22, 0 , 0, …., 123, 0, 12/8/2016 ]
Land Cover
Classification Model
Model
Multi-Dimensional
Dimension varies
i1
i2
i3
in
Image (i1, i2, i3, … , in)
b1
b2
b3
bn
Image (b1, b2, b3, … , bn) Object (reference parcel or holding LPIS) Land Cover Class System
Crop Code & Crop Description
1.1.1 xxxxxxxxxxxx1.1.1
Segment
(Polygon: x1y1, x2y2, x3y3, x4y4)
PHASE 2: Descriptor Extraction for each object per Image
Feature Extraction t22
All pixels
All Bands or/ and Indices
SIFT
SURF
ORB
CHIST
Method(s)
[ 0, 0, 0, 32, 0 , 0, …., 123, 0 ]
Descriptor (scaling)
PHASE 1: Indices (i) extraction using atmospherically corrected bands (b)
PHASE 3: Model Creation (data driven)
Train set
Input (time series data for polygons) Output( Labe)l
[ 0, 0, 0, 22, 0, 5, …., 100, 0, 12/04/2016] [1.1.1]
[ 0, 0, 0, 32, 0, 0, …., 123, 0, 07/05/2016] [1.1.1]
…….
[ 4, 1, 1, 22, 0, 0, …., 123, 0, 22/10/2016] [1.1.1]
Training
SVM
RF
NN
Evaluation
PHASE 4: Crop estimation for other parcels using the data-driven input -output model(s)
Input (time series data for a polygon)
[ 0, 0, 0, 22, 0 , 5, …., 100, 0, 12/8/2016 ]
…….
[ 4, 1, 1, 22, 0 , 0, …., 123, 0, 12/8/2016 ]
Land Cover
Classification Model
Model
Multi-Dimensional
Dimension varies
i1
i2
i3
in
Image (i1, i2, i3, … , in)
b1
b2
b3
bn
Image (b1, b2, b3, … , bn) Object (reference parcel or holding LPIS) Land Cover Class System
Crop Code & Crop Description
1.1.1 xxxxxxxxxxxx1.1.1
Segment
(Polygon: x1y1, x2y2, x3y3, x4y4)
PHASE 2: Descriptor Extraction for each object per Image
Feature Extraction t22
All pixels
All Bands or/ and Indices
SIFT
SURF
ORB
CHIST
Method(s)
[ 0, 0, 0, 32, 0 , 0, …., 123, 0 ]
Descriptor (scaling)
PHASE 1: Indices (i) extraction using atmospherically corrected bands (b)
PHASE 3: Model Creation (data driven)
Train set
Input (time series data for polygons) Output( Labe)l
[ 0, 0, 0, 22, 0, 5, …., 100, 0, 12/04/2016] [1.1.1]
[ 0, 0, 0, 32, 0, 0, …., 123, 0, 07/05/2016] [1.1.1]
…….
[ 4, 1, 1, 22, 0, 0, …., 123, 0, 22/10/2016] [1.1.1]
Training
SVM
RF
NN
Evaluation
PHASE 4: Crop estimation for other parcels using the data-driven input -output model(s)
Input (time series data for a polygon)
[ 0, 0, 0, 22, 0 , 5, …., 100, 0, 12/8/2016 ]
…….
[ 4, 1, 1, 22, 0 , 0, …., 123, 0, 12/8/2016 ]
Land Cover
Classification Model
Model
Multi-Dimensional
Dimension varies
i1
i2
i3
in
Image (i1, i2, i3, … , in)
11. TRAIN SET
Land Cover Classification using Indices
Crop Cover (ha) Parcels No.
Wheat 545 498
Stone fruits 367 364
Legumes 158 142
Maize 229 164
Trees 244 210
Fallow 354 358
Pasture 1,976 80
Total 3,873 1,816
Total Area (ha): 282,600
Agricultural Areas (ha): 53,580
12. TEST SET
Land Cover Classification using Indices
Crop Acc. (%)
Wheat 90%
Stone fruits 81%
Legumes 87%
Maize 57%
Forest Trees 61%
Fallow 65%
Pasture 78%
Total 74%
Crop Cover (ha) Parcels No.
Wheat 2,164 4611
Stone fruits 2,400 5045
Legumes 1,344 1588
Maize 2,917 3823
Forest Trees 549 894
Fallow 3,258 8744
Pasture 1,167 51
Total 13,799 25,264
13. Two back-end web services that rely on EO-S2 data:
• Polygon, time period -> Times series of indices, Estimated Crop Pattern
• Polygon, time period -> Crop classification
GAIA Smart Farm & GAIA Subsidy (CAP)
Advisor
Farmer
GAIA Cloud
API
That allows:
• On-the-fly automated cross checks of parcels declaration (proactive control)
• Full scale automatic compliance cross checks (post-declaration control - locate the
outliers)
• Crop monitoring – Smart farming (Advices based on Facts)
Plus Smart Farming data:
• Digital data (Valid Train set).
• Other type of data describing soil and atmospheric conditions (better mapping).
14. Next steps:
• Period 2017 – 2018: Subsidy and Smart Farming services for 16 large scale
areas (pilot sites) and for 15 significant crops.
• Collaboration for specific sites and crops with the partners of DataBio H2020
project (3 years from 2017 to 2020).
• Enrich the modelling process with more Descriptors (SL1 + L8).
• Deep Learning (for specific time windows using VHR data).
• Big Data Analytics (Outliers)
Thoughts - Discussions :
• IACS will take advantage of other Cloud Platforms (Food Security).
15. Thank you for your attention
Any questions?
15
PanagiotisILIAS(NP)
p_ilias@neuropublic.gr
DimitrisKAPNIAS(GAIA)
d_kapnias@c-gaia.gr
22nd CAP/IACS conference, 24-25 November 2016, Lisbon - Portugal
Editor's Notes
What is GAIA-Business S.A?
GAIA-Business S.A. is a coalition between the farmers, the banking sector and the industry, aiming to the growth and transforming of Agriculture in Greece.
GAIA EPICHEIREIN activities?
GAIA EPICHEIREIN is certified - Farmer Service Offices (FSOs) - by the Greek Payment Agency until 2020 to coordinate the collection process, perform various cross compliance checks and develop and apply correction services.
GAIA EPICHEIREIN advisory services?
GAIA EPICHEIREIN also provides to the Greek farmers advisory, through the GAIA Cloud Platform.
What is GAIA Subsidy?
The GAIA Subsidy provide of advisory services to the farmers related to the two pillars of CAP aiming to promote the main objectives of CAP (2014-2020)
aiming to simplify the whole process (including the declaration process) maximizing the effectiveness of the EU budget.
What is GAIA Smart Farm?
GAIA Smart Farm allow the farmers to make data-informed decisions regarding their cultivation practices. These services focus on the precise and timely application of inputs, such as fertilizers, pesticides and irrigation.
What are the main data sources? For both services (Subsidy – Smart Farming)
Generally speaking there are 5 main data sources.
And our core business is to collect the data and transform them to facts and finally to advices (How, When, What, Why).
What are the similarities between Subsidy – Smart Farming?
Both services make use of similar cloud services and take in advance the same data sources.
In a generic perspective view IACS is subset of the farming data (parcel limits, farmer id
Both services trying to support the Farmer using data.
Why it is possible to provide advices to farmers for large scale agricultural areas?
The IoT and Cloud Infrastructure allows the monitoring of large scale agricultural areas and the provision advisory services to the farmer related to pest management, irrigation, fertilization and subsidy through an integrate.
What is GAIA Sense?
GAIA Sense is a network of GAIAtron telemetric stations, which are designed and built by NEUROPUBLIC for collecting soil and atmospheric data through the integrated sensors and for controlling actuators (e.g. for automating irrigation).
The Role of Sentinel Data?
Sentinels provide open, sort revisit (means high collections frequency), high resolution, multispectral data.
The Role of Sentinel in CAP implementation?
Sentinel can been used online during the declaration process, as part of a cross-check that analyses the data to examine the correctness of the declaration and provide other valuable information in terms of compliance. It can be used either to support the farmer’s decision making or provide, if need, conclusive proof to the paying agencies. The farmer can use the GAIA Subsidy services through the GAIA platform, not only to provide ancillary data relative to the control process, but also to validate its compliance during the declaration process.
Through the use of Sentinel data, GAIA Subsidy services enriched the declaration process and adopted a more pre-active approach that follows the amendments of EU regulations (809/2014, 640/2014) for preliminary checks and increase of efficiency.
This approach also promotes the use of data coming from other Food Security platforms, additionally to IACS data, for the agricultural policy monitoring.
Gaia Observation integrates Remote sensing with Image Processing, Machine Learning, Spatial Modeling and Data Analytics in a cloud platform.
Offering:
More data
More info
Higher spatial and time resolution
Phenological phases detection?
And to combine vegetation's indices with the phenological stages of each crop
To calculate the the NDWI from Sentinel-2, we’ll be using band 8A as the NIR band, and band 11 as the SWIR band.
Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI)
soil adjusted vegetation index (SAVI),
green soil adjusted vegetation index (GSAVI),