Semantic Web Enabled Smart Farming

18
Semantic Web Enabled Smart Farming Semantic Machine Learning and Linked Open Data Application for Agricultural and Environmental Informatics CSIRO COMPUTATIONAL INFORMATICS Raj Gaire | Research Software Engineer 22 October 2013 IN COLLABORATION WITH

description

Slides from my talk at 1st International Conference on Semantic Machine Learning and Linked Open Data (SML2OD) for Agriculture and Environmental Informatics

Transcript of Semantic Web Enabled Smart Farming

Page 1: Semantic Web Enabled Smart Farming

Semantic Web Enabled Smart FarmingSemantic Machine Learning and Linked Open Data Application for Agricultural and Environmental Informatics

CSIRO COMPUTATIONAL INFORMATICS

Raj Gaire | Research Software Engineer

22 October 2013

IN COLLABORATION WITH

Page 2: Semantic Web Enabled Smart Farming

Smart Farm

• Informed Farming• Precision agriculture

– Sensors, information system, decision support systems

– System exists within a farm-gate

• Connected Farm• Devices in the farm are connected with each other and the world using

internet

• Farmers are connected to the farm devices, other farmers and experts

• Things (e.g. Cattle) in the farm can be monitored remotely.

• Integrated Farm• Includes Farmers in the supply chain - suppliers, logistics, consumers – back

to the farmers to complete the loop.

Presentation title | Presenter name2 |

Page 3: Semantic Web Enabled Smart Farming

Kirby ‘Smart’ Farm

• Location Armidale, NSW, Australia

• Farm Area: 739 Hectares (or 1827 Acres)

• Smartfarm Area: 269 Hectares (or 665 Acres)

• Livestock: Cattle, Sheep

• Devices: 100 Soil Sensors

2 Weather Stations

Cattle ear tags

Flex, Alix PC, 3G Modem etc.

Presentation title | Presenter name3 |

Page 4: Semantic Web Enabled Smart Farming

Presentation title | Presenter name4 |

Page 5: Semantic Web Enabled Smart Farming

What do farmers want?

• Measurement data produced by 100 sensor every couple of minutes?

• Weather measurement produced every couple of minutes?

• Cattle location updated frequently?

• Farmers are interested in the alerts about the things in the farm.• Cattle leave the farm

• When to sow

• Current market value of their livestock

• Soil in a paddock is compacted

• Researchers/Experts are interested in the data.

Presentation title | Presenter name5 |

Page 6: Semantic Web Enabled Smart Farming

Our Architecture

Presentation title | Presenter name6 |

Page 7: Semantic Web Enabled Smart Farming

Smartfarm Ontology

Presentation title | Presenter name7 |

Page 8: Semantic Web Enabled Smart Farming

Data Dimensions

Presentation title | Presenter name8 |

Page 9: Semantic Web Enabled Smart Farming

GSN Extended

• Geo-Spatial Analysis• Implemented using R and Java packages

• Event (Alert) Processing• Extended GSN to process event descriptions and produce alerts

• Synchronous and Asynchronous events

• Farms can create their own events

• Semantic Web Enablement• Sensor data stored in MySQL

• Linked data are produced using defined URIs

• Statistical data are stored in Virtuoso triple store

– Provides open access to everyone, analyse data using SPARQL

– VisualBox and Google APIs for visualisation

Presentation title | Presenter name9 |

Page 10: Semantic Web Enabled Smart Farming

Event Detection

Presentation title | Presenter name10 |

Web Form… …. …. ….… …. …. ….… . Submit

Event Manager

Event Description

Storage

Event Evaluator

Event VirtualSensor

Message Queue

GSN Storage

Event Description

Alerts

Page 11: Semantic Web Enabled Smart Farming

Important Links

PURPOSE LINK

Homepage (ROOT) http://smartfarm-ict.it.csiro.au

Semantics http://smartfarm-ict.it.csiro.au/semantics.jsp

Latest Data http://smartfarm-ict.it.csiro.au/latest

Specific Latest Data ROOT/dataset/sensornets/kirby-farm/type/{id} [/latest

Time Series Data Cube ROOT/dataset/sensornets/kirby-farm/{type}/{id} [/year/{year}/[month/{month}/[day/{day}/[hour/{hour}]]]]

VisualBox Home http://kirbyfarm-virtuoso.dyn.dhs.org/visualization/

SPARQL endpoint http://kirbyfarm-virtuoso.dyn.dhs.org:8890/sparql

Presentation title | Presenter name11 |

Page 12: Semantic Web Enabled Smart Farming

Presentation title | Presenter name12 |

Page 13: Semantic Web Enabled Smart Farming

Presentation title | Presenter name13 |

Page 14: Semantic Web Enabled Smart Farming

Visualisation

Presentation title | Presenter name14 |

Page 15: Semantic Web Enabled Smart Farming

Future Works

• SPARQL based access to dynamically generated data cubes

• Machine Learning over the Data

• Integrate satellite data

• Social Farming

Presentation title | Presenter name15 |

Page 16: Semantic Web Enabled Smart Farming

Machine Learning Opportunities

• Cost of Sensor Networks

• Variations are possibly correlated and predictable• Soil variation, elevation -> soil ec, temp, vwc

• BOM forecast -> farm weather

• Data collected over last 2 years • Use to generate predictive model

• Produce sensor data without sensors.

Presentation title | Presenter name16 |

Because data from Sensor networks in farms worth more than the sensor networks!

Page 17: Semantic Web Enabled Smart Farming

Acknowledgement

Kerry Taylor

Laurent Lefort

Michael Compton

David Henry

Ali Salehi

David Lamb

Gregory Falzon

Derek Schneider

Ashley Saint

Presentation title | Presenter name17 |

Page 18: Semantic Web Enabled Smart Farming

Computational InformaticsRaj GaireResearch Software Engineer

t +61 2 6216 7090e [email protected] www.csiro.au/CCI

CSIRO COMPUTATIONAL INFORMATICS

Thank you