From travelling a two-way road to navigating across scales Claudia Bauzer Medeiros Institute of Computing, UNICAMP Brazil
efarms and NavScales projects supported by FAPESP MICROSOFT RESEARCH VIRTUAL INSTITUTE, (2007-2010) and (2012-2014) And in-between
Outline efarms - context efarms - results and beyond NavScales context NavScales results so far
efarms a two-way road from small farms to the networked world 2007-2010
Outreach of Cooxupé
Satellite sensors Cooperative COOXUPE Images Maps Ground sensor network Reports Models Databases UNICAMP PUC-Campinas Farmers Produce
Satellite sensors Two-way Road Cooperative COOXUPE Images Maps Ground sensor network Reports Models Databases UNICAMP PUC-Campinas Farmers Produce
Expertise involved UNICAMP Computer Science (Databases) (Image Processing) (Compilers and computer architecture) (Wireless networks) UNICAMP Agriculture Engineering (Agricultural model development) (Soil Sciences) (Remote Sensing) One (very) large coffee cooperative (COOXUPE) 14,000 farms Three states in Brazil
GOALS Science & Technology Sensor-based agricultural data management Models, methods, algorithms Data management, integration, analysis, visualization Low cost wireless data communications in rural areas => Infrastructure
GOALS Social Impact Decision support in agriculture Models and software tools Wireless data communications in rural areas => Low cost digital inclusion
RESULTS
Result overview Training in multidisciplinary research Financed by project - 4 undergrad, 1 MSc Involved Agriculture 3 PhD, 1 MSc Involved CS 4PhD, 3 MSc, 6 undergrad Publications 7 journal papers 18 conference papers 1 poster, 1 short paper Technological results Low cost networks Prototypes (software and hardware) Crop yield models
Main issues Sensor callibration and low cost networks Models for yield forecasting Image processing (with Genetic Programming) Databases and interoperability Time series mining Content annotation Sensor data management Escience infrastructure
Results (Low cost) Networks From sensors to Internet
Satellite sensors Cooperative COOXUPE Images Maps Ground sensor network Reports Models Databases UNICAMP PUC-Campinas Farmers Produce
Phase 1 Lab Tests
Phase 2-3 - Internet
Phase 4 3 farms in 2009 Cooperative Moisture network Radiation, temperature
UNICAMP
Phase 4 - Sensor 3 m pole (4 sensors)
Sensors and cornstalks
Visible sensor
Daylight Jan 22, 2010 Underground Humidity (2 levels) Daylight Temperature (air, soil)???
Daylight Jan 22, 2010 Underground Humidity (2 levels) Daylight Temperature (air, soil) Two-way road
Phase 5 Real Farms & the 2-way road
Results Models, Images, DB
Models
Spatio temporal evolution How does this vary? What does it contain? Forecast? Comparison with others? Similar patterns? WHAT TO PLANT WHERE, WHEN AND HOW? 31/10000
32/10000
Thanks to Dr Rubens Lamparelli, UNICAMP
Models NDVI NDVI 0,8 0,7 0,6 0,5 Sugar cane 0,4 0,3 0,2 0 3 6 9 12 15 18 21 24 27 30 33 36 Decades Annotations 0,8 0,7 0,6 Urban area NDVI 0,5 0,4 0,3 0,2 0 3 6 9 12 15 18 21 24 27 30 33 36 Decades 0,8 0,7 Annual culture Image Processing Series Mining NDVI 0,6 0,5 0,4 0,3 0,2 0 3 6 9 12 15 18 21 24 27 30 33 36 Decades
Image processing
COFFEE IDENTIFICATION (Genetic programming)
Segment and classify
Series mining
Input 2 nd Closest 3 rd closest Farthest
Input 2 nd Closest 3 rd closest Farthest
Annotations
Workflow-driven annotations
A Case Study Get_ Annotatio n Schema Get Provenan ce Data County Classify, segment Obtain Yield Valida - <metadata> - <extendinfo> - <location> <county>http://www.lis.ic.unicamp.br/~carla/agriczoning.owl#municipio_350950</county> <state>http://www.lis.ic.unicamp.br/~carla/agriczoning.owl#estado_35</state> <country>http://www.lis.ic.unicamp.br/~carla/agriczoning.owl#pais_55</country> </location> - <product> <name>http://www.lis.ic.unicamp.br/~carla/agriczoning.owl#coffee</name> </product> </extendinfo> </metadata> 43
Interoperability
Satellite sensors Cooperative COOXUPE Images Maps Ground sensor network Reports Models Databases UNICAMP PUC-Campinas Farmers Produce
Interoperability architecture
Some lessons Multidisciplinary projects need long-time cooperation Deployment of sensor networks need LOTS of expertise in telecommunications and electronics Economically viable nice solutions are not necessarily adopted
Some lessons Multidisciplinary projects need long-time cooperation Deployment of sensor networks need LOTS of expertise in telecommunications and electronics Economically viable nice solutions are not necessarily adopted Nice projects never end
efarms Offspring a) Sensors + event processing b) Time series annotations c) NavScales project
Interoperability Architecture - part of PhD /FAPESP Ivo Koga
Time series annotation (+CNPTIA/EMBRAPA Mapagri project) Felipe Silva
Serial annotator
NavScales Navigating through scales in space, time and domain knowledge 2012-2014
Goals Help Scientists to Navigate across data in multiple scales Navigate across literature in multiple domains Share, reuse and collaborate
Team Computer Science Geo-spatial databases Semantics Compilers / course dissemination Institute of Biology Bio-acoustics Faculty of Agriculture Yield models and planning
Satellite sensors Multi-scale scenario Ground sensor network 56
Fonoteca Neotropical Jacques Vieillard
Paper EDBT 2013 workshop
Management of integrity constraints for multi-scale geospatial data FAPESP MSc defended March 2013 Joao Longo
Fonoteca Neotropical Jacques Vieillard
**** METADATA ****
FAPESP-funded PhD research Daniel Cugler
Curation of metadata
Work being developed at U of Minnesota with Shashi Shekhar
Multi-focus issues Multidisciplinarity and multi-expertise Needs, models, requirements, vocabularies Heterogeneity Of teams, data, procedures, models, methods Complementary ways of looking at the world Complementary ways of approaching a problem computationally
From travelling a two-way road to navigating across scales NICE PROJECTS NEVER END
MORE TO COME
Acknowledgements FAPESP and Microsoft Research Virtual Institute UNICAMP IC, FEAGRI, CEPAGRI, IB STUDENTS Colleagues from USP CNPq and CAPES COOXUPÉ EMBRAPA - CNPTIA