Reveal R&D: Testing Photogrammetry for Improving Utility Mapping in International Cities

Mapping utilities in dense urban environments requires precise and up-to-date information about the above-ground infrastructure (such as manhole lids, curb lines, and stormwater grates), as these features can provide important clues to the location of underground utilities. 

At Reveal, we have found that high-resolution imagery of above-ground features can be difficult to obtain, especially when mapping densely populated cities. Where aerial imagery of the street level is available, it is often obscured by the surrounding high rises or covered by overhanging structures, or lost in the cast shadows of tall buildings. When alternative sources of surface data are available, they are often out of date, incomplete or overpriced.  

To solve this issue, Reveal is testing new tools to develop an efficient way to collect above-ground data that is accurate, up-to-date and cost-effective. This study explores the feasibility, accuracy, and efficiency of using photogrammetry to create detailed geospatial images of urban roads and sidewalks to improve our utility mapping processes. 

Current Methods of Collecting Above Ground Utility Data 

One of the first steps in many utility locating activities is a process called Desktop reconnaissance, where a team member will review any existing plans, documents and maps that relate to an area of interest. Where geospatial aerial imagery is available, Reveal performs a process called Picking, where a software application is used to draw around the relevant surface features, with this data then being used to inform on-site work and often included in customer deliverables. 

Google Earth TM has revolutionized how we see cities around the world, enabling us to view any city in the world aerially and from the street with a click of a button. However, in cities with many high-rise buildings, it may be physically impossible to achieve high-resolution imagery due to the altitude requirements of the drone or reconnaissance aircraft. 

It is possible to use Google StreetView TM to investigate anomalies up close and supplement missing data from as-builts, but using this data for geo-referencing violates Google’s terms of service, rendering it impractical for utility mapping purposes. Plus, the geospatial accuracy of the StreetView TM and Google EarthTM data is never explicitly stated and tends to vary from location to location. 

In New Zealand, Land Information New Zealand (LINZ) provides high-quality, publicly available geospatial maps, which makes aerial picking of utilities more accessible. However, these maps are expensive to produce and are updated infrequently, and if a street happens to be covered by a canopy or scaffolding or shadow, the utility features may be obscured. Picking utilities requires clear high-resolution images of associated utility features, such as manhole covers, drains, and excavation scars along the roads to accurately geo-reference their locations and verify as-built information.   

Surveying Above-Ground Infrastructure  

Several tools from the surveying industry allow for on the street surveying of above-ground utility infrastructure, including a traditional Total Station for Topo surveying and mobile LIDAR scanners. 

A Total Station is an electronic optical instrument used for surveying and building construction that utilises an electronic transit theodolite integrated with electronic distance measurement (EDM) to perform topographic surveys of project sites. 

A Topographic Survey, also known as a detailed survey or ‘Topo survey’, delineates the position, boundaries, and characteristics of natural and built features within an area. The survey produces a precisely measured map of the areas existing topography, with detailed labelling and classification of surveyed features; these are commonly undertaken in anticipation of engineering or construction ventures. 

The ultimate survey output may show contour lines, perimeter demarcations, natural attributes, structures, pavements, street signs, visible utility infrastructure, and various other features, depending on the nature of the project. 

Mobile LiDAR scanners can capture data at high speeds, allowing for efficient mapping of large areas such as roadways, parking areas, or streetscapes. These scanners emit laser beams into the environment and measure the time it takes for the light to bounce back once it has hit an object. Mobile LiDAR systems can emit thousands of laser pulses per second, allowing them to quickly capture detailed spatial data. It provides highly accurate topographic survey results that can be used for 3D visualizations and modelling to inform projects on above-ground features. 

While these tools are effective for creating topographic surveys that can aid in documenting the above ground infrastructure, compared to using photogrammetry, there are significant drawbacks in cost, expertise and time expenditure. 

As Reveal expands its international presence, we benefit from higher-quality, and more up-to-date ‘Reality Capture’ methods to ensure the accuracy and accessibility of data captured by third parties is sufficient to allow us to maximize the value of our desktop workflows; this in turn, increases the cost-effectiveness and efficiency of any on-site activities for us and our customers.  

With this in mind, Reveal is testing a variety of photogrammetry solutions to obtain the specific data required for precise surface feature mapping, as for us, using topographic surveys is ‘better than nothing’ rather than the ideal. 

The Photogrammetry Approach 

Photogrammetry is a form of ‘Reality Capture’ that calculates three-dimensional information from a series of two-dimensional images, allowing for the generation of accurate 3D reconstructions of real-world scenes and objects. Photogrammetric scans can be used to create accurate, high-resolution data assets that record the shape, location, and condition of above-ground features present on the street surface.  

The goal of our testing is to collect photogrammetric data of roads and sidewalks in order to create high-definition geospatial imagery and other derivatives. The imagery collection process would need to be performed by international onsite teams and would not require extensive training or equipment costs.  

Table 1:  A comparison between above-ground surveying tools 

Reveal is working with the Wellington based computer vision startup, Sensori, who uses a novel combination of cloud-based photogrammetric algorithms and film-grade cameras to create 3D snapshots of building projects at a fraction of the cost of LIDAR solutions.  
Sensori, have mainly focused on hand-held data collection within construction sites, whereas, for our application, data collection needs to be performed with the camera mounted to a vehicle.  

The method we are testing with photogrammetry uses a commercial off the shelf (COTS) box camera with a fisheye lens, a standard survey-grade GPS radio, and a suitable vehicle all of which are easily accessible in most cities around the world at relatively low cost compared to drone-based surveys, fixed-wing aerial surveys or terrestrial LIDAR surveys.  

To collect images the camera is secured on a vehicle and driven at 30km/h through a survey area, while capturing multiple frames per second of the surrounding street. When the street is densely populated by traffic blocking the view of sidewalk features, the camera can be placed in a backpack and walked down the sidewalk to survey the areas that was obscured during the vehicle-based survey. 

Once images have been collected the data is uploaded into Sensori’s cloud platform for processing delivering a point cloud mesh and geospatial imagery within hours of the survey being performed.  One of the advantages of Sensori’s solution is that where geospatial accuracy of the resultant artefacts is not to a high enough standard, it can be retroactively corrected by on-site surveys and then reprocessed.  

The results of a test survey can be seen in the following images:

Image 1: A test survey site on the Kapiti Coast, New Zealand where a team member captured data walking the sidewalk, with a camera secured to a backpack.  

Image 2: The photogrammetric image delivered once processed in the Sensori Online viewer.

Image 3: The resultant point cloud displayed in a point-cloud viewer.

The geospatial images created by Sensori’s software allow our team to pick street features straight from the viewer, the collection of which was fast and low cost, requiring few expertise from the field team in data collection. In the future, there is also the opportunity to apply AI techniques to high-resolution photogrammetric scans to automatically identify and categorize features. While this approach is not the only one Reveal are testing, the results so far have been promising. 

Follow the Reveal LinkedIn page to keep up-to-date on Reveal research and development.

#ResearchAndDevelopment #Photogrammergy #Innovation

.