Identification of innovations for TMR to access timely, efficient, accurate, and sustainable data and insights to support state-wide planning and decision-making, and enable future focused smart, connected and autonomous infrastructure, networks and vehicles.
Geospatial data, particularly remotely sensed data, are pivotal in the development of smart, connected and autonomous road networks across Australia. Specifically, they facilitate insights relevant to the planning and maintenance of road networks. This project was framed by four key inter-related areas of:
- Demonstrations of novel approaches of using AI/ML to extract information and insights about what’s on the road, around the road and in the road corridor from very high-resolution imagery products. How many quality derivative datasets can be extracted from remotely-sensed datasets to augment TMR’s existing transport network data in a reliable, timely and automated way?
- Demonstrations of novel approaches to the creation of derivative products trained on very high-resolution data which can be applied to generate these same derivate products from lower-resolution aerial photography data. How effectively can the extraction of information and insights be scaled up about what’s on the road, around the road and in the road corridor for a larger area, and how good is the data?
- Research-based investigation into the accuracy and precision limits when using AI/ML algorithms to extract enriched road information and insights at different image resolutions. When derivate products are made from different source datasets that have different ages and accuracies, how can we have confidence they can work together effectively? How well can we quantify how different data extraction methodologies (including training AI/ML models for lower-resolution imagery), and image resolutions affect the accuracy and utility of derived datasets for specific use cases?
- Demonstrations of novel applications of AI/ML to mine non-visual (i.e., multispectral, hyperspectral, thermal, LiDAR/point cloud, radar, other) remote sensed data to detect information and insights about the roads, and road corridor. Can we detect problematic unplanned changes due to natural disasters, vegetation growth, drainage issues, subsurface changes before they become a major problem?
This innovation labs project would not have been possible without the challenge ownership and funding provided by the Queensland Department of Transport and Main Roads. The expertise in the geospatial future needs of the transport sector was instrumental in being able to select meaningful, useful capability demonstrators to undertake the four challenge projects for this program.
The quality outputs of these challenge projects is testament to the capability demonstrators who delivered these four projects – Mapizy, Veris, the Australian Road Research Board (ARRB) and the University of Wollongong (UoW).
The TMR Spatial Labs program 2022 created a challenge-based approach to industry innovation and capability. Using FrontierSI as engagement, management & procurement mechanism, the program focused on machine learning (ML) and artificial intelligence (AI) analytics to extract insights from remotely sensed data to address challenges and gaps in the TMR road network. Four $50,000 projects of 3-4 months were undertaken, framed by an overall vision and TMR provided data (drone, vehicle mounted, and other imagery data collected by TMR):
- Road Analytics in Multiresolution Aerial Images (Mapizy) – This project explored the potential of artificial intelligence and machine learning models to create derivative analytics about the road and road network (lane markings, divider and traffic islands, features from the surrounding environment such as trees and buildings) from 10-12cm resolution aerial imagery and 30-50cm satellite imagery.
- Seeing in the dark: Early subsurface anomaly detection (Veris) – This project utilised 3D Radar technology for sub-surface mapping of a pilot area of TMR road to detect early signs of road degradation below the surface, such as moisture, pavement layers and services.
- Hazardous location identification for head-on collisions with roadside assets and objects: A deep learning approach (Australian Road Research Board) – This project uses Digital Video Road (DVR) and artificial intelligence tools to automate the detection trees with a trunk diameter greater than 10 cm which are around the road corridor (ranging from closer than 1 m to up to 10 m from the road), i.e., ‘killer trees’.
- Developing Deep Learning Tools for Road Information Extraction from Georeferenced High-resolution Imagery (University of Wollongong) – This project aims to use georeferenced image/video TMR-collected data and artificial intelligence, computer vision, and deep learning methods to create a suite of five tools to automate detection of different types of road defects, traffic signs, road features, road centrelines and solid/dashed lane lines.
The TRM Spatial Labs program was designed to support innovative projects within the broader context of the strategic objectives of the TMR Vision to create a single integrated transport network accessible to everyone. The TMR Spatial Labs initiative makes a small, but critical contribution to the TMR Strategic Plan (2019-2023) helps build strategic partnerships, support a balanced R&D portfolio, help secure IP, demonstrate new technologies, and accelerate capability and entrepreneurship across the Queensland transport, AI and spatial sectors.
Following the experience of facilitating and managing the TMR Spatial Labs 2022 program, FrontierSI has seen first-hand the energy, engagement, technical potential and spark of innovation that these projects have created. Project participants found tangible value in engaging with TMR data to explore novel approaches to using AI/ML and spatial data analytics to improve TMR’s ability to access timely, efficient, accurate, and sustainable data and insights that support state-wide planning and decision-making, and enable future focused smart, connected and autonomous infrastructure, networks and vehicles.
ARRB is honoured to be engaged in the TMR Spatial Lab 2022 program. Working with FrontierSI was significantly efficient, starting from the inception meeting, where the team members were introduced and the project plan was confirmed. Once we agreed on the deliverables, FrontierSI organised fortnightly meetings so all the partners could monitor the progress and provide feedback. FrontierSI was also proactive in helping us to access the required data for our project. As a result, we did not face delays due to the lack of communication. We are pleased to work with FrontierSI on this project and look forward to future collaborations. – ARRB
This project was well managed with the FrontierSI team being involved. With the support from a team of excellent communicators and tech leaders, we kicked off the project very quickly knowing that every data we need can be sourced with ease. During the project and the reporting phase, we had the privilege of being able to contact the team when needed and discuss the findings and seek feedback. All in all, this new type of working with the government through FrontierSI helped to successfully finish the project with a better experience working on similar projects. – Mapizy
To learn more, contact FrontierSI at email@example.com or connect with Project Manager, Roshni Sharma, at firstname.lastname@example.org.