An Extensible and Integrated Software Architecture for Data analysis ...

An Extensible and Integrated Software Architecture for Data
analysis and Visualization in Precision Agriculture
Li Tan 1,2 Ronald Haley 1 Riley Wortman 1 Qin Zhang 2
1
School of Electrical Engineering and Computer Science, WSU
2
Center for Precision and Automated Agricultural Systems, WSU

Outline
• Introduction and motivations
• Research questions on information retrieval and reuse in
context of precision agriculture.
• System overview
• Data importation and integration.
• Data-flow-driven design for data processing and visualization.
• Client support for desktop and mobile applications.
• Conclusions

IT in Precision Agriculture
Information technology is crucial to all the three stages in precision
agriculture:
• Collect data from a variety of sources:
• Sensor networks, vehicle-based data collection, and Aerial and satellite
images.
Research Question: how to import data from different sources with different
format
• Analyze and visualize the data.
• Analysis flow varies depending on actual applications, e.g., operations
characteristics.
Research Question: How to build a reconfigurable analysis flow?
• Visualize the data to support decision-making.
Research Question: How to visualize the data so farmers can understand?
• Use the data to optimize field operations.
Research Question: How to deliver the data and result to a variety of devices
for on-the-fly mobile access?

Our solution
Research Question I: how to import data from different sources
with different format
Solution:
1. Maintain data importation as a separate and customizable module;
2. A meta-data model in XSD for a user to define input data format.
Research Question II: How to build a reconfigurable analysis flow?
Solution: Build a data-flow driven and model-based analysis subsystem.
1. User may define their own workflow;
2. User may define their own operators.
Research Question: How to visualize the data so farmers can understand?
Solution:
Build a data-flow driven and model-based analysis subsystem.
1. User may define their own workflow;
2. User may define their own operators.

Our solution
Research Question IV: How to visualize the data so farmers can
understand.
Solution: Provide multi-feature 3-D plot for multi-dimension data sets. A
plot may contain the following information:
1. Color map;
2. Multiple layers;
3. 3-D contour.
Research Question V: How to deliver the data and result to a variety of
devices for on-the-fly mobile access?
Solution: Build a client-server architecture to support a variety of
computing devices, including,
1. Desktops;
2. Ipad App.

Data Importation and Integration
• Data processed in precision farming has large
variation in format.
• Data comes from various sources: remote sensor
array, vehicle based data collector, etc.
• Data format is often decided by manufacturer of
devices.
• Our solution: create a flexible data importation
module supporting user defined format in XML
• The allowed XML formats are defined as a metadata-model
in XSD.

Meta-data model in XSD
XSD enables the data importation module to recognize
and validate data definition file in XML

? xml version =”1.0” encoding=”ISO− 8859− 1”?>
< DATAFILE fi l e n a m e =” d a t a par .csv”>
< uid />
< sensor array>
< sensor id=”1” X=”mm” Y=”mm” device=”PAR” / >
...
< sensor id=”8” X=”mm” Y=”mm” device=”PAR” / >
< / sensor array>
< time>
< date time>
< full date format=”YYYY/MM/DD” / >
< full t i m e f o r m a t =”HH:MM:SS . SS” / >
< /datetime>
< /time>
< other data />
< / DATAFILE >
< DATAFILE filename=”data gps . csv”>
< uid />
< movement>
...
< / movement>
< coordinate>
< coordinate lon lat>
< longtitude unit=”degree”/ >
< latitude unit=”degree”/>
< / coordinate lon lat>
< / coordinate>
< / DATAFILE >
< /DATA>
Sample data definition
file in XML

Data-flow-driven design
for Data Processing and Visualization
• The workflow of data processing and visualization largely
depends on actual application and operational characteristics.
• Data processing and visualization supports the customization
through,
• Data-flow-driven design supporting graphical manipulation
of workflow by re-arranging the data flow and operators.
• Custom-defined data operators.
• User may implement his own data operator.
• Each operator has to have a well-defined interface
specifying input and output data format.

Data-flow-driven design

Data visualization
• A flexible data visualization module supports the
render of multi-dimension data sets.
• Multi-dimension data sets may be rendered,
• On the additional dimension (Z-axis);
• As multiple layers of images;
• Using customized color map.
• Data may be overlaid on a terrain map to provide
geological-referencing.

Geo-referenced data visualization

3-D visualization
3-D visualization of Photosynthetically Active Radiation (PAR) data.

Client-Server Architecture Design
• Supported clients include desktops and iPad;
• Enable the on-the-fly access on mobile devices (iPad).

Implementation
• Server built on the open-source data-mining tool
rapidMiner.
• Implemented in Java.
• Desktop client to communicate with the server and
handle user-computer interaction.
• Data visualization is implemented using jzy3D, an
openGL java library.
• A Ipad client is developed using iOS development
kit.

Conclusion
A software architecture for data analysis and visualization in
precision agriculture with the following features:
• A data-importation module supporting custom-defined data
input format through meta-data-model in XSD;
• A data-flow-driven design for reconfigurable data processing.
• Dataflow may be re-arranged;
• Support custom-defined data processing operators.
• A visualization module supporting:
• Geo-referenced data visualization;
• Visualizing multi-dimension data set.
• A client-server architecture supporting:
• Mobile devices for on-the-fly data access and visualization

Thank You!
Questions and Comments?