Many years ago, I read an interesting article in Survey Review on using a 35mm camera to capture enough information to provide measurements for accident forensics. This was long before digital cameras, with their JPEG & RAW files, and it was a fairly crude form of terrestrial photogrammetry. However, it was horses for courses, results being adequate for the situation, plus quick and cheap. Since the advent of laser scanning, it seems photogrammetry has taken somewhat of a back seat.
Having spent 3 days in late October covering the Leica HDS 2009 conference, I was interested to find out what was new and exciting in the world of terrestrial photogrammetry. I found it a couple of weeks later during my first ever WebEx experience, when I hooked up with other 3D heritage enthusiasts in a 3-continent Skype call. (WebEx Communications is a Cisco company that provides on-demand collaboration, online meeting, and web and video conferencing apps). Organized by Adam P. Spring, ADAM Technologys UK rep for Educational & Cultural Heritage, it amazed me how far weve come interacting in real time. Joining Adam was Dr Caradoc Peters, head of the Dept of Archaeology at Plymouth Univ. From The Getty Institute in Santa Monica, CA, was Rand Eppich, and from Orinda, CA, Justin Barton with CyArk (the non-profit organization set up by Ben Kacyrathe father of laser scanning). Jason Birch, managing director of ADAM Technology, Western Australia, provided the substance in demonstrating, via Powerpoint, what could be done with the right software in turning a DSLR (digital single lens reflex) camera into a viable competitor to laser scanning. Although Jason stressed that much could be done with even a basic DSLR, they usually use a semi-pro Canon 5D with fixed lenses–mostly 50mm, 100mm & 200mm (zoom lenses seldom allow replication of focal length every time). Calibration is done on all cameras.
There are some basic differences between the two:
Each point in a laser scanner point cloud is an independent sample while the pixels in an image have a fixed relationship with respect to each other. So, you can georeference photogrammetric data using a small number of known points, confident that the rest of the data will be correct relative to those points; you can never be sure the same is true for scanned data.
When it comes to range, photogrammetry gives a great deal of flexibility to the user. E.g., changing to a lens with double the focal length allows an image to be captured from twice as far away with very little effect on the results.
Accuracy of 3D data for a DSLR depends on the relationship between lens focal length and distance, and the base-to-distance ratio. The user is free to operate from whatever distance is most convenient, simply choosing the lens and ratio to fit the accuracy requirements of the job. Unlike laser scanners, planimetric accuracy is typically more accurate than the depth accuracy (usually other way around for laser scanners).
Range using a DSLR surpasses that of a scanner and can be more accurate.
Regarding that last point, Jason noted that BHP Billiton Iron Ore routinely use distances up to 1200m, and as many as 20 images (using fans) being captured from each camera location. This minimizes the number of different locations required to capture a typical pit wall. And, the Escondida mine in Chile has been mapped with a 3cm pixel size on the ground from 2.8km. This technique allows extremely large areas to be captured and modelled in 3D very quickly with minimal fieldwork. If you think about it, this is terrestrial photogrammetry, or aerial mapping at 90?, on a stable ground platform. No wobbly planes yawing and pitching. Although ADAM Tech will often put the DSLR on a tripod, it can be handheld. Its all in the geometry, and the trick is sufficient overlapping photos, plus, of course, proprietary software3DM Analyst–that provides an automatic least squares best fit.
To see how quickly a project can be done, Jason took an open pit mine, covering 400m 280m. From the panel below, you can see that using a 100m lens takes about 10 mins capture time, 13 mins to process, and generates 2 million points. Youd be hard pressed to do even one scanner set up in that time.
One use that has literally been making waves is for the offshore oil & gas industry. Oil rigs are incredibly expensive to run, and being out of action for even a brief time eats into profits. Steel legs and anchors deteriorate in salt water and harsh storm conditions, and need to be monitored for excessive wear. In the past, divers used videounsafe and not accurate. You cannot take a scanner down a few hundred metres; besides, where would you set up the tripod? But, a DSLR, mounted on a mini-sub, can take numerous overlapping photos. Then, using the software, chain link wear can be determined, and as built modelling performed for subsequent modifications.
In a vastly different application, one of their customers–a dentist–published a paper in a peer-reviewed journal showing he had achieved accuracies of 15 microns using a pair of 6MP DSLRs and the software.
It gets better. If terrestrial doesnt quite hack it, and you need to capture part of a site from overhead, say of stockpiles, theres always the UAV (Unmanned Aerial Vehicle). A tiny helicopter, half the size of a person, and equipped with a DSLR, is remotely controlled to fly up to 400 feet above ground and take a series of overlapping photos. For one stockpile 175 points/m2 (15 million total) were collected, with a point accuracy of 20 mm. Field time was 30 mins (6 mins in the air), and processing time 30 mins. A scanner that had been doing the work took 3.5 hrs and 3 hrs respectively. Volume comparisons agreed within 0.3%.
At $50K for commercial ventures, and $10K for non-profits carrying out heritage work, the software is not cheap. The UAV will also run you $40K. But consider this: unlike laser scanning where equipment at $100K is the high cost item, plus upgrades at similar prices, terrestrial photogrammetry costs are in the software. And, for about $3K, using a good DSLR plus suitable lens, youre off and running.
For further information, you can visit their web site at www.AdamTech.com.au. I can also recommend Jasons comprehensive Powerpoint presentation, which covers all you want to know about practical terrestrial photogrammetry, the geometry, camera calibration, how to shoot various types of sites, and use the software. I also suggest getting hold of his paper, Laser Scanning vs Photogrammetry, which covers pros and cons of each.
Note: This article appeared in the March/April 2010 issue of Geomatics World.
About the Author
Nick Day, FRICS, FRGS, PLS, is retired from the California Department of Transportation (Caltrans). He can be reached at feasibility.nick@gmail.com
The 8.75Mb PowerPoint is HERE
A 732Kb PDF of the paper Laser Scanning vs Photogrammetry is HERE
You can download a 10.8Mb PDF of the updated version of this paper HERE
A 2.97Mb PDF of the article as it appeared in Geomatics World is HERE