Thursday, July 31, 2014

VRI Hall_ VR LiDAR Model

Above is our 'Field Model' of the VRI Hall and its immediate surroundings that we constructed from our LiDAR scans.

The terrestrial scanner owned by RMIT Universities School of Architecture and Design, was used to survey the site in readiness to prepare and finalise construction drawings for the future design transformations to the hall and its adjacent tennis court area, into a lively, inspired and creative learning and community space, and as a part of the Neighbourhood House in Traralgon.

The model above is only a small part of the total point resolution that we gathered from 6 separate scans, and we thought that rather than post the raw scan point model (images below), or our surface reconstructed model, that it would be nice to experiment just a little.

VRI Hall_ VR Exterior Photogrammetry Model

Above is a VR (virtual reality) model, where you can interactively navigate around the exterior of the VRI Hall model that was constructed using photogrammetry.

The model therefore, has been entirely constructed using only photographs that we took while onsite and waiting for the LiDAR scanning to finish. We believe the results aren't too bad, considering that no post remodelling at all has been done. This is clearly visible in the lack of photographic coverage to the roof areas, and what looks like snow.

Apart from the more obvious issues of the model (namely some strange surfaces), it has served as a great test, and a very quick 3D mapping / study.

VRI Hall_VR Interior Photogrammetry Model

Much like the previous post, instead this is a capturing of the interior.
A total of 43 images were use to construct the dense mesh model. What you are seeing above is a decimated mesh so as not to lag the VR navigation too much.

We have been working to prepare a comparative post on the LiDAR scans that we conducted

Friday, July 25, 2014

VRI Hall LiDAR and Photogrammetry Scanning

On Wednesday 23rd we drove down to Traralgon to commence the scanning and surveying of the VRI hall and its surrounding using both a LiDAR system and some photogrammetry techniques that we have been discussing in posts prior to this one. 

We will be displaying some of the images and information collected on these two techniques in future posts.

Above: The rear of the hall

Above: The rear of the hall, seen from the view of a quadcopter

Above: The west of the hall and old tennis courts, seen from the view of a quadcopter

Above: Scanning the interior using LiDAR.

About the larger project: 
The Victorian Railways Institute (VRI) Hall is a community based project for the Traralgon Neighbourhood House’s & Digital Shed project and the Reactivate Latrobe Valley project.

For project information on the VRI Hall, please visit it on our web site by CLICKING HERE

Monday, July 21, 2014

Photogrammetry Scan Test 7

Again, something completely different, we thought we would test a drone. Problem was that for a test we didn't want to bother some of the contacts that we have with drones, so we turned to the internet for some footage and found this video by MisterG of the Leaning Tower of Pisa to extract some of the stills to test.

Now our guess is... that his drone was carrying a Canon 5Dmkiii or similar, though it doesn't really matter as we extracted the images in After effects to 1080p resolution for the test.

The take away:
Thats more like it again! The only down fall is the coverage of the images to parts of the scene, and could be fixed with re projected textures in maya, zbrush or similar. It again proves that a aerial type image capture on a sunny day is far superior for a great starting point to an really quite detailed and textured urban or pastural model.

To still try:
Time for a real project study! TBC

Sunday, July 20, 2014

Photogrammetry Scan Test 6

Now for something completely different, we have changed the site to a residential one, and returned to a canon 700D. However this time, instead of landscape we have gone portrait in an effort to change two things.
01. that the entire building height be captured in the shot, including more foreground capture without going to a fish eye type lens, AKA GoPro
02. to insure that we have sufficient overlay of 60 to 70 percent.

The take away:
Thats more like it! The street has been done and quite a lot of detail has been produced. The downside is that the image set is quite large, consisting of somewhere in the range of 120 images that are spread across 3 cluster groupings. We also turned the matching accuracy up to high, which meant the processing time was about 5hrs for everything, including Textures and a decimated mesh production. Not that bad really when compared to some of the processing of LiDAR sets.

To still try:
What about a drone?

Saturday, July 19, 2014

Photogrammetry Scan Test 5

For this test we again used a iphone5.
Weather was the best we have had yet, so we raced out and took some shots from the top of the Hub down onto the construction site next door.

The take away:
Wow This is more like it! We only took 7 or so images and it looks great, this is going to be hard to improve on, and reconfirms the work that is being done by companies like aerometrex through one of their models of Melbourne as seen here . No doubt the elevated positioning of where we took the images from was a big factor here, and clearly shows the potential of photogrammetry when developing models at the urban block, city scale and above, as also demonstrated by NASA and their DTM models. This isn't to say that photogrammetry isn't great for object based scanning, because it is as seen here.

To still try:
Test iPhone on shaded conditions
Other locations
Different overlap percentages

Friday, July 18, 2014

Photogrammetry Scan Test 4

After the previous test, we decided to use the iphone5 under harder lighting conditions and by returning back to the Swanston Street side.

The take away:
Hmmm, a bit of a mixed bag. The results appear to be quite poor again compared to the sunny scenario. Still not too bad as a quick test, though it is now obvious that this side is problematic due to the shade, and if we intend (or have to) shooting under these conditions, we will need to premark and code the site.

To still try:
Other locations
Different overlap percentages

Thursday, July 17, 2014

Photogrammetry Scan Test 3

For this test we used a CanonIXUS700 as a point and shoot.
Weather wasn't ideal with the shadows, but at least it was sunny for a brief period.

The take away:
It appears to be the best test we have done yet, though we knew this would be the case with the improve in the weather.
The major down side with this test is there is still no foreground being constructed in the point cloud, and will need the images to possibly have more areas of interest in common here. The mesh does look cleaner though, and could still be cleaned up a lot more in post if required.
We also tried video capture with this camera, however the output was only 640x360 in its video mode and the result was not ideal. Photoscans own documentation suggests using a resolution of at least 720p as a bare minimum.

To try:
Other types of cameras (iPhone)
Other locations
Different overlap percentages
Different numbers of images in set (to reduce processing times)
Different resolutions and type of capture, i.e. Static photo sets, Timed intervals and Video to then extract static frames.

Wednesday, July 16, 2014

Photogrammetry Scan Test 2

For this test we used a GoPro Hero3 and corrected the lens distortion in Photoshop before processing.
Again the weather was far from ideal, with it raining as we were taking the images and therefore there were a lot of wet reflective and shiny surfaces.

The take away:
Well this was a worse result than the first test, and we don't really know why other than the obvious. What we suspect is, that lens of the GoPro looked slightly damaged on closer observation, and maybe the performance of the focus and DOF wasn't ideal in the conditions. We of course had no control over this with a GoPro, and would still think that it would be a ideal choice for photogrammetry due to its size, weight, cost and capture methods. Especially in Drones!!!!
We also found this work Click Here  by Jonathan Sileoni on his blog that proves that it should work well.

To try:
Other types of cameras (iPhone and a point and shoot)
Other locations
Different weather and surface conditions
Different overlap percentages (landscape vs portrait)
Different numbers of images in set (to reduce processing times)
Different resolutions and type of capture, i.e. Static photo sets, Timed intervals and Video to then extract static frames.

Tuesday, July 15, 2014

Photogrammetry Scan Test 1

For the first test we used a Canon 700D, and the weather was far from ideal, with it raining as we were taking the images, and therefore there were a lot of wet reflective and shiny surfaces.

But as good a place as any starting place!

We would also like to point out that we have not optimised any of the settings unless stated, or we forgot to mention it. Sorry
We are using Photoscan for the software and then have taken the model into Maya. Please be aware there are others choices in software such as the free 123D catch by Autodesk.

The take away:
Well considering it was less than ideal photography weather, the resulting 3d model produced isn't too bad. A good start especially considering, we would be able to use something like this for bash Viz level work and detailed massing / proportions .

To try:
Other types of cameras
Other locations
Different weather and surface conditions
Different overlap percentages
Different numbers of images in set (to reduce processing times)
Different resolutions and type of capture, i.e. Static photo sets, Timed intervals and Video to then extract static frames.

Photogrammetry versus LiDAR

Recently we have been conducting some further research on ways of effectively modelling, or extrapolating survey data from which to develop more accurate and detailed models of a build environment. Combined with this, we are also thinking of the possibilities for Landscape Architecture which could be further extended when coupled and integrating with similar fast approaches that have their origins in Matte Painting and Projection Models. In the past we have posted about, and on a number of occasions have used LiDAR ourselves, to gather a precise scan information sets of 3d data on a site, however we thought it was about time that we do some work with the ever more popular method of Photogrammetry. (that we actually post)

So what are the two and their differences?  
Well....     LiDAR is a remote sensing technology that measures distance by illuminating a target with a laser and analyzing the reflected light. (from wikipedia)
Where as Photogrammetry is the science of making measurements from photographs, especially for recovering the exact positions of surface points. Moreover, it may be used to recover the motion pathways of designated reference points located on any moving object, on its components and in the immediately adjacent environment. (from wikipedia)

It must be pointed out, that neither of these are new, and have been going through rapid development in the surveying disciplines in recent times. In fact, Photogrammetry was used before the Apollo missions to reconstruct and further understand, the surface of the moon prior to the luna landings (though mostly done by human calculations, rather than as it is today, with algorithmic software), and is still used widely being used today by NASA to model distant planets among other things.   Click Here to read more

Photogrammetry we believe, has massive potential due to it being a incredibly cheap alternative when compared to LiDAR. What was rather surprising when we first read it was that it can be as accurate, or indeed more accurate than LidAR if set up correctly with ground and camera positional data.

The others advantages (and there are many more) would be the cost affordability. This isn't just reflected in the software and hardware but the capabilities of light weight UAV Drones for their use in aerial surveying type applications instead of traditional aircraft.
Further advantages are the on site setup times, taking into account the actual scanning and processing are also heavily in favour of Photogrammetry over LiDAR.
So this is where we still felt we needed to test and document some of these capabilities for ourselves.

We will be detailing the tests that we have conducted soon in coming posts, and are just editing some quick films for each.

Two interesting articles we found comparing the two, if you are interested are:

Aerial Perspective: Photogrammetry Versus Lidar

Scanning the margins

Some Past Posts on this subject:
UAV mapping of antarctic
Modelling London