BIG DATA, A.I. AND
How reliable is your Data?
Production chains, government agencies and scientific researchers are experiencing a “Big Data Deluge” from all type of sources, sensors, space, legacy systems, etc. Users are exposed to novel concepts such as Internet of Things, Machine Learning and other buzzwords. But how safe and reliable is your Data coming from satellites or IoT devices? Are you actually making good sense of it?
Since 2013, Open Parallel have been contributing to the design of the compute platforms of the Square Kilometre Array radio-telescope (SKA), the largest mega-science project of the world. In the next decade, the SKA will process data and images at rates that are orders of magnitude greater than anything existent today. Their work is now moving into cyber security, open source operating systems, and large, large amounts of data that need to be processed and stored in a useful way. They are investigating applications of these technologies into New Zealand’s Primary Sector in areas such as Precision Agriculture.
This talk will discuss Data Interrogation Tools, UX designs and systems engineering techniques so users can access data from i.e. European Satellites with the same ease as from sensors at their farm, and combine them to make adequate decisions i.e. for Precision Irrigation and Water Management.
Nicolás Erdödy - CEO, Open Parallell Ltd
Nicolás Erdödy is founder and CEO of Open Parallel Ltd., a high-tech R&D firm. A recognised international speaker, Nicolás is behind Multicore World - a specialised high-tech conference with an economic development agenda for New Zealand.
He holds a Master of Entrepreneurship from the University of Otago, lives in Oamaru, in the South Island of New Zealand, and knows how to ask for a beer in five human languages.
Lauder, a hot spot for satellite climate observations
Over the last ten years, the small town of Lauder in Central Otago has become one of the most measured sites in the Southern Hemisphere by a progression of climate observing satellites. Lauder is so valuable to spacecraft operators and science teams because of its location on the globe, its local climatology, and for the vast array of ground truth measurements conducted at NIWA’s Lauder atmospheric research observatory. In this talk, Dave will describe why this barely-there Otago settlement is, for a group of international scientists, the major centre of New Zealand.
Dave Pollard - Atmospheric Scientist, NIWA
Dave has nearly twenty years’ experience in atmospheric remote sensing. After studying Physics with Space Science and Technology at the University of Leicester in the UK, he went on to spend ten years working for the UK Met Office as an Instrumentation Scientist conducting microwave remote sensing from the UK’s atmospheric research aircraft.
In 2014, he shifted to Lauder to work for NIWA as the principal investigator for the Total Carbon Column Observing Network (TCCON) site at Lauder.
Exploring the benefits of cloud computing services on satellite data processing
After developing nationwide hydrological datasets in the Smart Aquifer Characterisation (SAC) programme, GNS Science started to explore the current capability of cloud computing services for satellite data processing. The main reason for that was to gain speed in the very time-consuming calculation of nationwide or large-scale satellite data processing.
In the SAC research, these typically took months to download and process, therefore GNS Science wanted to see if cloud computing services could speed up these calculations. The results were very successful, so they decided to focus on the use of the Google Earth Engine (multi-petabytes archive in cloud computing service, including cloud-based machine learning algorithms) to analyse New Zealand’s wetland, flood plains, forestry and agriculture. This resulted in mapping wetlands and floods with a resolution of 10m x 10m (e.g. Otago flood in July 2017, Edgecumbe floods April 2017), and councils have also asked to analyse other land-use change events, such as:
- forest loss and gain over time per region and for the nation.
- classification of willows in wetlands
- classification of variety of vegetation in wetlands
- land use change in general (e.g. spraying)
As GNS Science was already using the Google Earth Engine, there was no need to download data, nor process them on their own computers. Hence, all the requests could be easily analysed in a a very short amount of time. Dr Westerhoff will present application demonstrations for multiple regional councils, as well as nationwide datasets.
Dr Rogier Westerhoff - Remote Sensing Scientist - Hydrogeology, GNS Science
Dr. Rogier Westerhoff is a remote sensing scientist working at the Hydrogeology Department of GNS Science. Rogier is mostly involved in the incorporation of satellite data into large-scale (nationwide) hydrogeological models that are able to fill in the gaps in data-sparse areas. Rogier works with the Google Earth Engine (GEE), a cloud computing service that facilitates fast processing of multi-petabyte archive of satellite data, from the 1980s to a few days old, from multiple satellites.
In the last year Rogier has assisted regional councils in their mapping of floods, vegetation, wetlands and forestry. The aim was to build capability to roll out these cloud-computing services, including machine learning, within the regional councils.
Artificial Intelligence, Machine Leaning, 'deep learning' models in New Zealand Agriculture and the problems that can be solved with satellite data
The science of Artificial Intelligence and Machine Learning made a significant breakthrough (“equivalent to the Fosbury Flop” in the high jump) in 2012 when a new “deep” model won a computer vision competition. That “deep learning” model rendered all previous models obsolete, revolutionising the way problems are solved. Applications of deep models being developed around the world include self-driving cars, speech recognition, computer assistants (Siri), real-time language translation, lip-reading, object classification in photographs, fraud detection, stock market prediction and marketing applications such as product recommendations and optimal advertisement delivery.
Precision AI creates deep algorithms and models to solve data science problems in New Zealand and worldwide. Applications developed include analysis of medical and agricultural images (ground, drone and satellite images), sensor analysis (including hyperspectral), pest detection and identification, pre-harvest yield prediction (counting buds and berries) and bioinformatics (breeding). They specialise in commercialising the application of deep learning and machine learning so believe this would provide a unique skill set to the speakers at the What on Earth Colloquium.
This presentation will cover:
- What is Artificial Intelligence, Machine Learning?
- What does “deep” mean for New Zealand agriculture?
- What is Precision AI currently working on and what problems need to be solved using satellite data?
Particular machine learning and deep learning applications using imagery Precision AI are working on with large commercial customers would be included i.e. (Waka Digital, Eurofins and Zespri). Deep learning is changing the way we think about agriculture and through collaboration, we will reach new heights.
James Beech - Senior Data Scientist, Precision AI
James Beech is a Senior Data Scientist at Precision Artificial Intelligence with over 15 years experience in software development, advanced analytics and data visualisation. His experience spans across financial services, telecommunications and the agricultural sectors. James specialises in open data, data infrastructures, business intelligence dashboards and predictive modelling. James has a particular interest in the application of big data through the use of statistical and analytic techniques to solve business problems. Technologies and toolsets: VB, VBA, SAS, SPSS, R, Tableau, PowerBI, Microsoft Azure cloud, Hadoop, Matlab.
Build automated feature extraction systems from data collected from any platform
Advancements in Earth observation data from satellites, planes and drones means that big data is a big part of our world. This paper presents Orbica’s research to build automated feature extraction systems from data collected from any platform. Orbica can turn this big data into big information by combining the best of geoprocessing with the latest advancements and methodologies in artificial intelligence (AI), to create methodologies that can extract features of the earth’s surface from only 3-band imagery. They currently have algorithms that can extract – with high accuracy and very fast performance – building outlines, roads and surface water types from this 3-band imagery.
Orbica's advancements include:
- Ability to consume data at high speed
- Ability to automate AI and geoprocessing to output raster and vector datasets
- Agnostic of dataset. It can be tuned to extract information from many resolutions
- Very scalable due to ground-up architecture of the models
- Levels of certainty are provided against all datasets through confusion matrixes
This presentation will appeal to all those interested in taking existing and future imagery data collections and turning these raw images into actionable information using the latest advancements in AI and geospatial.
Kurt Janssen - CEO/Founder, Orbica
Kurt is a geospatial entrepreneur and founder of Christchurch-based location data intelligence company Orbica. He has a BSc Hons (Geography) from the University of Canterbury and a wealth of experience working for central government and private companies in New Zealand and California, U.S. Kurt is passionate about geography and has developed a talented team of disruptive thinkers around him who apply their cumulative experience and knowledge to solve tough problems with cost-effective, location data enabled solutions. Orbica is leading the geospatial sector with its investment into artificial intelligence and its ability to automate traditional business workflows to improve accuracy, efficiency and value for clients.
Monitoring Big Issues with Big Data – developments at the University of Waikato
The Environmental Research Institute at the University of Waikato has been active in utilising satellite image archives to develop processes for monitoring water quality of lakes in New Zealand, and developing a model to predict ice melt in Antarctica from daily MODIS land surface temperature data. Thousands of images have been downloaded and analysed using automated processes. Different AI techniques have been compared to determine which is the most accurate in different contexts. A brief overview of these projects will be provided, as well as some innovations in combining GIS data and spectral data to identify individual Pohutukawa trees from space.
Dr Mathew Allan - Senior Researcher, Environmental Research Institute, University of Waikato
Born in Te Kuiti in 1979, Mathew grew up in Hawkes Bay and has a Ph.D from the University of Waikato.
Mathew’s Ph.D. thesis investigated the use of remote sensing and in situ data to create empirical and semi-analytical algorithms for the retrieval of chlorophyll a, suspended particles and water surface temperature in New Zealand lakes. He also investigated the use of spatially resolved statistical techniques for comparing satellite estimated and 3-D simulated water quality and temperature.
Mathew’s current area of research includes spatially explicit catchment modelling, ecologically coupled hydrodynamic modelling of lakes, water quality in relation to GIS, and the remote sensing of water quality.
Dr Lars Brabyn - Senior Lecturer in GIS, University of Waikato
Lars Brabyn is a senior lecturer in Geography at the University of Waikato specialising in GIS analysis and remote sensing. Over the years he has supervised many brilliant students who have made major contribution to remote sensing techniques and innovative applications.