Embedding Automated Spatial Data Integration in Street Surveys: A Bayesian Data Fusion Approach
Huanhuan Chen and Anthony Cohn
Statutory records of underground utility apparatus are notoriously inaccurate, so street surveys are usually undertaken before road excavation takes place to minimize the extent and duration of excavation and for health and safety reasons. This involves the use of sensors such as Ground Penetrating Radar (GPR). The GPR scans are then manually interpreted and combined with the expectations from the utility records and other data such as surveyed manholes. The task is complex owing to the difficulty in interpreting the sensor data, and the spatial complexity and extent of under street assets. We explore the application of AI techniques, in particular Bayesian data fusion (BDF), to automatically generate maps of buried apparatus. Hypotheses about the spatial location and direction of buried assets are extracted by identifying hyperbolae in the GPR scans. The spatial location of surveyed manholes provides further input to the algorithm, as well as the prior expectations from the statutory records. These three data sources are used to produce the most probable map of the buried assets. Experimental results on real and simulated data sets are presented.