Charles Sturt University students develop machine learning models for government flood agencies

Charles Sturt University Computational Imaging students who have undertaken work experience projects with the NSW Customer Service Department have developed trained machine learning models that can be deployed in information system software to generate maps of flooded regions.

The models were designed during the NSW floods.

“In 2021, DCS sponsored nine Charles Sturt University students to complete image processing projects using Amazon Web Services (AWS), an aerial imagery computing and project platform ( cloud-based API), said Dr. David Tien, senior lecturer in information technology.

Members of Charles Sturt’s student team were Adam Blewitt, Andrew Smith, Patrick Funnel, Cameron Nyberg, Darren Sheehan, Ian Blott, Regan Frank, Thomas Godfrey, and Michael Senkic.






Blewitt says the goal of the AWS Aerial Imagery Project was to automate the process of extracting geospatial information from floods using machine learning models, reducing the time needed to obtain this information by two days of work. to two hours, improving the accuracy of information and consolidating work that was duplicated across multiple government agencies.

Following a flood, space services can fly an aircraft over flood-affected regions to collect multi-band (RGB and near-infrared) aerial imagery.

These images are then passed to a specialist for processing, to stitch the images together and perform orthorectification, the process of transforming an image of a curved surface into a flat image suitable for maps and removing distortion from the movement of the aircraft. .

Blewitt explains that previously these images of the flooded region were then passed on to several agencies such as SES and the Department for Primary Industries (DPI) for their own use.

“Organizations would then manually review and extract geospatial information from these images, such as the boundaries and size of the flooded area. This required a lot of time and expertise and delayed the ability of departments to make informed decisions following a natural disaster,” Blewitt points out.

Charles Sturt’s student team used images provided by the Space Services of past floods in Hawkesbury-Nepean (March 2021), Brewarrina (April 2021) and Lower Clarence (March 2021) to train models of machine learning to perform semantic segmentation of aerial flood images.

Blewitt defines semantic segmentation as “the process of assigning each pixel in an image to one of several categories, with the ultimate goal of this project being to isolate all pixels in the deluge into a single category”.

“We experimented with several models, including convolutional neural networks, Gaussian mixture models, and complex decision trees, each using different algorithms and learning methods to perform the same semantic segmentation task.”

“The models can be used on any flood, provided the appropriate near-infrared aerial images have been taken.”

Executive Director of Space Services Narelle Underwood said the agency was delighted to be able to partner with Charles Sturt University to help students take on a project with real world applications.

“This collaboration provides students with access to our subject matter experts in a range of fields, with the results providing benefits for space services and the community,” concludes Underwood.

Sherry J. Basler