Part I of this article introduced the concept of using satellites for construction site monitoring, and the SCSI project initiated to research this concept further. Part II looks at the outputs of this research so far.
The study area for this project is located in the inner part of Dublin Bay on the east coast of
Ireland. The Alexandra Basin Redevelopment (ABR) Project is located in the approaches to Dublin Airport and 2.5km south west of a nature and bird sanctuary, limiting options for aerial or remotely piloted aircraft systems (RPAS) surveys. The ABR Project is the first part of the larger redevelopment of Dublin Port that forms part of Dublin Port Company’s Masterplan 2012-2040.
The output of the photogrammetric process was a point cloud coloured by elevation, as illustrated in Figure 3a, where blue represents low elevations (e.g., the ground) and red represents high elevations (e.g., building roofs). The synoptic potential of satellites as a mapping tool is reinforced in Figure 3b, as this displays the extent of data captured during a single acquisition with multiple potential applications. Despite a focus on a single test site, the images covered the whole of Dublin city and we were able to map this in 3D from a single satellite overpass. The high spatial resolution of the Pléiades satellite imagery results in approximately one point per pixel, i.e., four points per m2, a density not dissimilar to many conventional airborne light detection and ranging (LiDAR) platforms.
Five test buildings were then selected for validation, with the target structures distributed throughout the site, ensuring robust testing of the methodology (Figures 3c-g).
These buildings represented a selection of structures around the port, including historical buildings for preservation, storage buildings, and buildings scheduled for demolition. These structures range from approximately 30m to 125m in length, and heights range from approximately 3m to 12m. The dimensions (length/width/height) of each building were measured in the satellite-derived point cloud and subsequently compared with measurements from the validation data (Table 1). Five measurements of each plane were manually recorded in a computer-aided design (CAD) environment (15 measurements per structure) and a series of accuracy estimates created including mean absolute error (MAE) and root mean squared error (RMSE).
The average RMSE for length, width and height was approximately +/-1m. In future tests, taking additional measurements will allow us to spot or remove the outliers and then we can look at the MAE (RMSE amplifies and severely punishes large errors) – an average MAE of approximately +/-0.5m was achieved for measuring dimensions of many of the structures. This cannot come close to competing with traditional terrestrial survey or photogrammetric methods in terms of accuracy (and
doesn’t pretend to try), but it is important to remember that this was carried out from a desk, with no boots on the ground, was almost completely automated, included nearly every building in Dublin city as an afterthought, can be applied as easily to Dublin as to Durban, and can be repeated as often as required.
Satellite imagery is widely considered to be a tool suitable only for large-scale mapping of environmental variables; however, in this study we have demonstrated the potential of this technology for automated progress monitoring at the building level. We have highlighted the utility of earth observation (EO) satellites by generating a spatially referenced time series of maps, surface models and point clouds – all common datasets essential for any BIM workflow. The full study also looked at satellite radar (high resolution TerraSAR-x, bathymetry, dredging in the bay and time series for 4D BIM). The European Space Agency (ESA) awarded us five Pléiades and five
TerraSAR-x, captured over a number of months at four-week intervals; however, we have concentrated on the photogrammetric component only in these articles. A scientific journal paper is in reparation that will go into the methodology in more detail and quantify accuracies of the other outputs, and we will report back at a later date.
Research in this paper was supported by the Society of Chartered Surveyors Ireland Research Fund and Ordnance Survey Ireland’s Research Initiative. All satellite datasets were provided by the ESA under a Third Party Missions licence and AIRBUS copyright. The authors wish to thank Dublin Port Company and Future Analytics Consulting for the validation datasets provided.
Dr Conor Cahalane FSCSI, FRICS
Department of Geography, Maynooth University
GIS and EO Analyst at Mallon Technology
Doctoral candidate at Maynooth University
Dr Avril Behan FSCSI FRICS
Lecturer and Assistant Head of the School of Multidisciplinary Technologies, TUD
Stephen Purcell FSCSI, FRICS, MIPI
Director, Future Analytics Consulting Ltd
Eimear McNerney FSCSI, FRICS
GIS and Mapping Specialist, Planning and Asset Management, ESB.