This project investigates new approaches to the processing and analysis of dense 3D point cloud data in the context of geospatial intelligence requirements, with the ultimate aim being to identify, evaluate and implement improved feature recognition, extraction and modelling of natural and man-made features from dense 3D point clouds generated by both ranging systems and 3D Images determined through multi-image matching of high-resolution imagery. Project outcomes will facilitate a more automated semantic tagging and geometric description of objects as varied as buildings, small structures, vehicles and earthworks.
The project objectives comprise ongoing integration of developed software tools and procedures into a data processing system that has been delivered to support day-to-day operations.
Work on the project has produced a number of enhancements in the capabilities, levels of automation and operational flexibility of the software tools being developed. Updated versions are being regularly delivered. Specific areas of R&D include automatic calibration of digital cameras, especially those with lenses of long-focal and narrow fields of view; the ability to extract reliable 3D information from uncalibrated and/or unknown digital imaging sensors; and advances in automated network orientation and 3D object reconstruction from unstructured, multi-image configurations via new approaches to feature-based matching.
Prof Clive Fraser
Science Director Program 4.2
Dr Alex Lee
Australian Geospatial-Intelligence Organisation
Department of Defence