Vision Ai and 3D Urban Modeling

Artificially Intelligent Survey and 3D models Generated on Demand

Artificially intelligent (Ai) software seems to have permeated many levels of our lives. From smart algorithms that control the content we see online to driverless cars. Seemingly ubiquitous, Ai has infiltrated the Remote Sensing sphere.

Ai will now start to impact the architecture, construction and engineering sector (AEC).

Surveying and remote sensing information are required for all AEC projects. Ai collected remote sensing data will now impact 3D modeling users, like SketchUp and Revit.

Remote sensing

Remote sensing is the scanning of the earth by satellite or aircraft in order to obtain information about it (source Wikipedia). The collected remote sensing data, like aerials, DEM’s, DTM’s, contour maps, building footprints and heights (Figure 01) etc. serve as necessary starting points for orienting and understanding a project site.

Ground survey’s

Ground survey’s, manually collected by people in the field, is used alongside remote sensing data. The combined information yields 2D CAD base files (Figure 02) of existing conditions for both above ground and underground existing conditions, like utilities, curb and road layouts, site features like street lights, etc…

Most projects experience several weeks to months delay from a project start and when the site survey and remote data is available. This slows down the initial design process, forcing many companies to rely on third party vendors like Google Earth, to just get a basic aerial of a location. Drones, another disruptive technology, do fill the gap for many projects, allowing firms to obtain some of the data on their own (Figure 03). That still requires a person on site flying the drone then going back to the office and processing the data.

But to obtain a 2D cad file of existing conditions (figure 02) requires time and drafting labor to produce. It is a manual process regardless of who completes the task be it a surveyor, draftsperson or designer.

Computer Vision Ai

Artificial Intelligence, specifically those related to computer vision identification or vision ai, are revolutionizing this traditional paradigm described above. Computer Vision has been around for a while.

More commonly used for facial recognition software and easily accessible to the average coder wanting to develop apps utilizing computer vision; usually for identification of a product, item, location, animal, flora, and not least of all driverless car’s navigation. (Figure 04)

There are many companies now using vision ai to scan the world, resulting in Ai derived remote sensing and surveying data produced quickly and for substantially reduced costs. Common examples include the updated data on Google Earth (and maps) and Bings online portals.

One company, Ecopia Tech (Figure 05) is at the fore front of this revolution. Using high resolution aerials, Ecopia has trained smart vision ai software to identify, classify and segment site features on an aerial. Not just outlines or blobs but accurate lines defining almost all the features found in the built world: Curbs, gutters, street lights, man holes, building footprints, parking areas, driveways, trails, ped lights, etc..

The process is simple to explain. Ecopia takes an aerial with good quality (7cm resolution is ideal), uses their trained ai algorithms which then creates outlines, layers and useful 2D CAD and 3D modeling line work, areas and delineations (Figure 06 and Figure 07)

Ecopia has been training these smart vision ai detection systems since 2009. The first step was training the computer to identify all the various site features. Now Ecopia is in the process of developing data sets for specific locations including the entire US, Canada, Western Europe Australia and more. Some locations are already available on demand. Other locations require a one month turn around. But in the coming future the process of creating coverage specific data will include whole swaths of the globe (think like Google Earth). The AEC industry, real estate, cities and counties, corporations and more will have this data at their fingertips for a fraction of the cost.

Currently, SketchUp and Revit users can obtain this data through PlaceMaker (www.suplacemaker.com) who serves as both reseller and file converter for end users. Full disclosure, the author is one of the owners.

Through PlaceMaker, Ecopia provides several output formats useful for AEC. The provided Ai generated survey can be easily converted into SketchUp, DWG or similar formats (this is what PlaceMaker does).

Unprecedented

For the first time in almost 4000 years, the human is removed from the drafting process of a site survey. The manual drafting process is replaced by an automated intelligent system. This saves huge costs.

Consider, Ecopia provides 1 square KM of data for $1,000 USD. A manual survey done, with people on the ground and using manual digital drafting would cost hundreds of thousands of dollars if not more for the same area.

The animation below is an example of what this data looks like in SketchUp and Revit. 3D massing models (OpenStreetMap buildings also imported using PlaceMaker) combined with the Ecopia site data is in the closest thing to an instant site model of existing conditions available. In the video 3D massing models from PlaceMakers OSM data set are included. All the roads, open spaces, driveway, parking, bridges, open space, park space, i.e. a fully classified articulated 3D model of existing locations.

More than just survey

Ecopia visions ai does not stop with a general survey of a site. Currently Ecopia is creating building volumes with accurate heights and articulation. The goal is to further texture these buildings into more accurate representations of reality.

And upon request, the aerial data can be further parsed to specific parameters. Down to the manhole, pedestrian light, tree and traffic light (Figure 09).

This is just the beginning. What comes next over the next five years will quietly revolutionize yjr way AEC project can be started and delivered through a project life cycle. Consider that companies like Google and Microsoft already have this level of accurate data for their use. Accessing this data from big tech is almost impossible to obtain. Companies like Ecopia and Nearmap are paving the way for easier access to accurate information that is also updated frequently enough, keeping it relevant.

The costs, the coming on-demand availability (and in some locations, already available data) will allow SketchUp and Revit users to quickly build detailed accurate models for design.

Daniel Tal, ASLA