A new trend is emerging in the geospatial industry: the use of photogrammetry techniques to generate realistic and accurate virtual 3D models of the environment. The art and science of photogrammetry has been around for decades but it is only in recent years that this technology has found prominence in the geospatial industry. What was just a concept a few years ago is now becoming more mainstream.
The photogrammetric process
Photogrammetry is the science of extracting accurate geometric properties of objects from overlapping photographs. The theory behind the photogrammetric process is as old as the invention of photography itself, however the widespread development of computers has drastically improved the photogrammetric workflow through automation.
From a geospatial point of view, aerotriangulated stereo images are the base product from which a number of 2D and 3D mapping products are derived, such as orthomosaics and digital terrain/surface models.
The latest technological developments such as high-resolution oblique camera imaging systems, parallel processing and analysis capability, cloud-based storage capacity and fast distribution networks have had a direct impact in the improvement of the photogrammetric process. The dense-matching, meshing and texturing pipeline makes heavy use of newer GPUs and parallel processing to deal with increased overlap between aerial and ground-level photographs, and increased spatial resolutions.
aero3Dpro is an example of a new 3D-modeling service aimed at producing geographically accurate, photorealistic and high-resolution 3D models of urban and naturals environment.
Dealing with complexity
3D photogrammetry is providing a new alternative to aerial Lidar products with faster and more cost-effective solutions and the potential to get a point density up to the single pixel resolution. Outside the undeniable advantages of LiDAR for vegetation penetration and power line detection, photogrammetry offers more flexibility in terms of acquisition and scalability as well as the possibility to generate photorealistic and geographically accurate visualization products.
Acquiring and processing large amounts of data are no longer an issue. The challenge resides in integrating and interacting with these large datasets. Geospatial and visualization software companies are slowly but surely developing new ways to integrate these new 3D models and ease the transition from 2.5D to a more intuitive 3D virtual world.
Currently, the mesh of complex 3D models has to be simplified without losing too much information in order to be integrated into existing interactive GIS and visualization software.
Some innovative companies such as Euclideon have developed new rendering techniques based on their Unlimited Detail technology. Through their Geoverse software, large datasets generated from ground-laser scanner and high-resolution photogrammetry are converted, compressed and can be instantly streamed to the user, eliminating the need for dedicated 3D cards, powerful CPUs and vast amount of RAM.
New applications based on 3D models are emerging every day from within the geospatial industry and beyond. The type of applications will vary depending mainly on the acquisition technique, the resolution of the photographs and the density of point-clouds used.
Based on current experience, two distinct developments are evident:
Firstly, a growing demand to provide complete 3D models, which includes all elements of a scene for visualization and animation. What was before a long and costly manual process done by professional modelers can now be greatly reduced. These new models can be used for public consultation, internal and external shareholder consultation and multimedia projects. The demand comes from real-estate, animation, advertising, the mining industry, and local and state government departments.
Secondly, a growing need to provide survey-grade 3D data for civil engineering and infrastructure projects. It didn’t take long for civil engineering companies to realize the exceptional potential of 3D models from photogrammetry to assist them on the design, rehabilitation and monitoring of critical infrastructure projects.
Potential simulation applications also range from flood, solar potential and telecommunication analysis, line of sight to flight simulation and training scenarios for military and emergency services.
3D photogrammetry in the mainstream
Google has largely been responsible for the democratization of GIS by initiating the transition between a geospatial, specialist-dominated community of a few hundred thousand to the general public with millions of users and growing. Other big players like Apple and Microsoft are also heavily involved in providing free and easily accessible online mapping content.
In recent years, these companies have focused on a more 3D-oriented content for their millions of users. The latest innovations from Apple, with its C3 technology, and Google, with its 3D imagery, make use of 3D photogrammetry techniques to automatically generate entire cities in 3D. Microsoft has also announced that it will provide automatically generated 3D content using Nokia HERE 3D maps (both 3D street-level and aerial). These aerial models are still rough and lack accuracy and resolution, which restricts their use to entertainment and visualization purposes, but have the advantage to cover large areas and provide a more immersive and realistic experience for their users.
These giants are setting a new standard in terms of online geospatial content and it looks as if 3D models generated from photogrammetry will become the norm in the near future.
Outside the geospatial world, democratization of 3D data is taking place with a growing number of online companies providing free services to generate 3D models from ground–level photographs. The most popular is 123D Catch from Autodesk, where users can upload a set of photographs of any objects taken from their mobile phone or still camera and generate a fully textured 3D model ready for 3D printing.
Toward a virtual 3D world
With a serious move toward the development of a 3D cadastre by the public and private sector, there is a growing need to have both the physical and legal three dimensional extent of individual parcel and associate interests. High resolution 3D models generated from photogrammetry are a key element in this initiative, not only at a building level but also to put this 3D cadastre in context by linking external assets such as trees, roads, poles and any other natural objects.
3D cadastre will become a successful management and decision making tool when cultural, decisional and technical challenges to integrate different disciplines, such as building information management (BIM), geographic information systems (GIS) and 3D visualization and simulation systems, are overcome. 3D models generated from photogrammetry are a key element to bridge the gap between these last two disciplines.
All the pieces of the puzzle already exist and, through standardization and interdisciplinary collaboration, there is no doubt that a robust solution will emerge where analysis can be performed based on the attributes of 3D elements.
In this three-dimensional world, creating an information-rich virtual 3D environment is an obvious path to take for the geospatial community to improve the way decisions are made and complex information is communicated. 3D photogrammetry has a major role to play in this initiative with the possibility to quickly generate, at an affordable cost, photorealistic and geographically accurate 3D models of urban and natural environments. Once the challenge of integrating and standardizing geospatial databases, complex 3D models and building information management data into a common physical and virtual platform is solved, the new geospatial revolution will truly begin.
Figure 1. Example of manually designed 3D building integrated into an aero3dpro 3D model of Adelaide, Australia generated from photogrammetry for real estate advertising
Figure 2: High-resolution Virtual City Model of Adelaide, Australia generated from aero3dpro
Figure 3. High-resolution natural environment aero3dpro model (including mesh on the right)
Figure 4. aero3dpro model displayed in Google Earth (top) and new Google Earth 3D imagery (bottom)
Figure 5. Example of feature capture based on an aero3dpro model