Automated Spatial Information Generation

ASIG A4 Handout (PDF)

Click here to go to the CRCSI-only page of documents and resources …

Over the past decade or so we have witnessed the emergence of several new automated data acquisition technologies within the spatial information industry. Under the broad classification of ‘imaging sensors’, such technologies include digital aerial photography, high-resolution satellite imaging systems, spaceborne and airborne interferometric synthetic aperture radar (InSAR) and airborne laser scanning (lidar). Moreover, such new technologies have not been restricted to aerial and space platforms, as we are seeing broader application of terrestrial imaging and laser scanning within mainstream surveying, mapping and GIS. A common feature of all these data acquisition systems is that they produce a wealth of primary data, but this is generally in a form that requires extensive further processing in order to generate information products.

The key to future productivity gains, which are necessary to keep pace with the ever increasing volume of spatial data and demands for temporal attributes of spatial information to be accommodated in GIS, is automation of the feature extraction process.

Research Themes

The program will have three themes:

• Research & development required to support the application of new data acquisition systems for feature extraction in spatial information. This includes metric sensor modelling, calibration, sensor orientation and geopositioning and development of the necessary mathematical models and computational processes needed to generate the data products from which cartographic features are extracted. A illustrative example in this context is high-resolution satellite imagery, which has been the focus of projects 2.1 and 2.11 in the original CRCSI. Here, it was necessary to develop the models and algorithms for sub-pixel image orientation and geopositioning before semi-automated feature extraction could be deemed feasible.
• Fundamental research into automated feature extraction from fused data obtained with multisensor systems, the focus being upon imaging and ranging data from digital aerial and satellite imagery, multi- and hyperspectral imagers, lidar and space- and airborne InSAR. Research will focus upon the development of increased levels of automation for robust and reliable extraction of both man-made features and vegetation by investigating object model-based, knowledge-based and trainable machine learning approaches of understanding image content.
• Applied research embracing the development of feature extraction algorithms, tools and processes for accurate and reliable feature database building in end-user specified application domains in the areas of defence mapping, utility asset management, urban planning and modelling, natural resource management and GIS in general. Specific applications areas identified include automated feature extraction to support topographic mapping from airborne InSAR data, tropical forest area monitoring and biomass estimation from combined radar and multispectral space imagery, and combined lidar and multispectral airborne imagery from UAVs to facilitate power line corridor monitoring and mapping, especially the automatic determination of vegetation clearance which is so crucial in bushfire prevention.