The underwater forest of Lake Tovel

As I wrote in a previous post, since 2005 Arc-Team is supporting Prof. Tiziano Camagna and Andrea Forti in their exploration of the underwater forest located in Tovel Lake (in Trentino - Italy). This project gave us some interesting data to test different techniques for underwater archeology, especially in the documentation field (like the use of SfM and IBM in extreme conditions).

I already reported some results of our experiments in this post and today I would like to publish the complete video we used to extract the 3d geometries of a flooded tree, thanks to the courtesy of Prof. Camagna, who kindly gave me the original record.

To help English reader, I translate here the superimposed text in the initial and final sequences:

"The underwater forest of Tovel Lake" (title)

"A landslide, as claimed by the studies conducted in 1992 by Oetheimer, caused the damming of the then envoy in 1597-1598, resulting in rising waters at the present level, submerging the forest located in the northeast of the lake" (intro)

"A photograph of the early '900 shows a tree of considerable size emerging from the water surface. This video shows the same tree today still firmly planted with its roots at the bottom of the lake." (intro)

"The under water forest of Tovel Lake;

people who participated in the making of the video:

Tiziano Camagna, Andrea Forti, Nicola Maganzini, Samuele Sozzi, Nadia Morani and Arc-Team (Cles) - Luca Bezzi and Alessandro Bezzi.

Video recorded with gopro hero 2.

Diving equipment.

Drysuits: Santi, Parisi; Computers: VR3 Delta P, Sunto; Lights: Scubatech, Fa&Me; Regulators: Apeks TXT 200, Interspiro Divator; Scuter: Suex; 

Gas blender: Tiziano Camagna, Andrea Forti

Special thanks to: the Municipality of Tuenno and the Adamello-Brenta Park

Video recorded in 2012" (credits) 

Currently Prof. Camagna started again with the exploration, while we will join him at the end of September (after +Alessandro Bezzi will return from the archaeological mission in Armenia). Soon we will post news and other reports regarding the project. 

Have a nice day!

From drone-aerial pictures to DEM and ORTHOPHOTO: the case of Caldonazzo's castle

Hi all,
I would like to present the results we obtain in the Caldonazzo's castle project. Caldonazzo is a touristic village in Trentino (North Italy), famous for its lake and its mountains. Few people know about the medieval castle (XII-XIII century) whose tower is actually the arms of the town. Since 2006, the ruins are subject to a valorization project by the Soprintendenza Archeologica di Trento (dott.ssa Nicoletta Pisu). As Arc-Team we participated in the project with archaeological field work, historical study, digital documentation (SFM/IBM) and 3D modeling.
In this first post i will speak about the 3D documentation, the aerial photography campaign and the data elaboration.

1) The 3D documentation 

One of the final aims of the project will be the virtual reconstruction of the castle. To achieve that goal we need (as starting point) an accurate 3D model of the ruins and a DEM of the hill. The first model was realized in just two days of field-work and four days of computer-work (most of the time without a direct contribution of the human operator). The castle's walls were documented using Computer Vision (Structure from Motion and Image-Based Modeling); we use Pyhon Photogrammetry Toolbox to elaborate 350 pictures (Nikon D5000) divided in 12 groups (external walls, tower-inside, tower-outside, palace walls, fireplace, ...).

The different point clouds were rectified thanks to some ground control point. Using a Trimble 5700 GPS the GCPs were connected to the Universal Transverse Mercator coordinate system. The rectification process was lead by GRASS GIS using the Ply Importer Add-on.

To avoid some problems encountered using universal coordinate system in mesh editing software, we preferred, in this first step, to work just with only three numbers before the dot.

2) The aerial photography campaign 

After walls documentation we started a new campaign to acquire the data needed for modeling the surface of the hill (DEM) where the ruins lie. The best solution to take zenithal pictures was to pilot an electric drone equipped whit a video platform. Thank to Walter Gilli, an expert pilot and builder of aerial vehicles, we had the possibility to use two DIY drones (an hexacopter and a xcopter) mounting Naza DJI technology (Naza-M V2 control platform).

Both the drones had a video platform. The hexacopter mount a Sony Nex-7; the xcopter a GoPro HD Hero3. The table below shows the differences between the two cameras.

As you can see the Sony Nex-7 was the best choice: it has a big sensor size, an high image resolution and a perfect focal lenght (16mm digital = 24 mm compare to a 35mm film). The unique disadvantage is the greater weight and dimension than the GoPro, that's why we mounted the Sony on an hexacopter (more propellers = more lifting capability). The main problem of the GoPro is the ultra-wide-angle of the lens that distorts the reality in the border of the pictures.
The flight plan (image below) allowed to take zenithal pictures of the entire surface of the hill (one day of field-work).

The best 48 images were processed by Python Photogrammetry Toolbox (one day of computer-work). The image below shows the camera position in the upper part, the point cloud, the mesh and the texture in the lower part.

At first the point cloud of the hill was rectified to the same local coordinate system of the walls' point cloud. The gaps of the zenithal view were filled by the point clouds realized on the ground (image below).

After the data acquisition and data elaboration phases, we sent the final 3D model to Cicero Moraes to start the virtual reconstruction phase.


3) The Orthophoto

The orthophoto was realized using the texture of the SFM's 3D model. We exported out from MeshLab an high quality orthogonal image of the top view which we just rectified using the Georeferencer plugin of QuantumGIS.
As experiment we tried also to rectified an original picture using the same method and the same GCPs. The image below shows the difference between the two images. As you can see the orthophoto matches very well with the data of the GPS (red lines and red crosses), while the original picture has some discrepancies in the left part (the area most far away from the drone position, which was zenithal on the tower's ruin).

4) The DEM

The DEM was realized importing (and rectifying) the point cloud of the hill inside GRASS 7.0svn using the Ply Importer Add-on. The text file containing the transformation's info was built using the relatives coordinates extracted from Cloud Compare (Point list picking tool) and the UTM coordinates of the GPS' GCPs.

After data importing, we use the v.surf.rst command (Regularized spline tension) to transform the point cloud into a surface (DEM). The images below show the final result in 2D and 3D visualization.

Finally we imported the orthophoto into GRASS.

That's all.

Extreme SfM: underwater archaeology

Hi all,

It is long that Alessandro and me wanted to write this post, but for one reason or another, we could not work on it. 

Today I decided to do it in order to answer two questions that people often asked us during conference or lessons:

1) Is it possible to work with SfM/IBM techniques underwater?

2) Is it possible to extract 3D from a movie?

As regards the first question, I can report that since we (Arc-Team) started to work with SfM and IBM (2009), we did also tests underwater and they gave us positive results. This is one of the main reason why we invested so much time on the research in this filed: SfM and IBM methodology, until now, is one of the best solution in archeology, due to its versatility (it can be used for underwater or aerial documentation, in low light conditions or in precarious situation, during mission abroad, etc...). We already underlined this concept when, with the help of Nicolò Dell'unto (Lund University), we compared different methodology to record 3D documentation of archaeological artifacts. The result of this experiment was presented during the ArcheoFOSS 2012, in Napoli (see the related slides and this post). During the workshop "Low cost 3D: sensori, algoritmi e applicazioni", we had the opportunity to better analyze the use of SfM/IBM in extreme working conditions, strengthening our point of view about this methodology (see the related slides). 
The image below is an example of an aerial 3D documentation done with an open source UAV and Python Photogrammetry ToolboX...

Aerial documentation with a KKopter and ArcheOS (PPT)

... while this other image shows the results we achieved using some pictures that Victor Jansa, of TUWA ("Tauchverein für Unterwasserarchäologie"), sent us to do a test.

Test with Victor Jansa's pictures (done with PPT)

In order to answer the second question, I can say that, facing our experience, it is possible to reconstruct 3D models from videos (and I guess this is one of the aims of SfM itself). We did some tests about this topic, getting acceptable result (at least regarding our primary target, which was to have a fast 3D object for further modeling operations). As an example, I can report one of the last post of Cicero Moraes, who used SfM from a youtube clip to get a 3D skull for forensic facial reconstruction aims. The image below is taken from Cicero's photo album:

3D skull obtained with SfM techniques from a movie

For a better explanation of what I wrote above, I think it is worth to show the results of a project we are undertaking since 2005 (trying to support Prof. Tiziano Camagna on his exploration of Tovel's Lake, in Trentino).  During this project we did several surveys, diving in different parts of the lake and especially in the South-West area, where lies a forest which is now underwater. In 2012 Tiziano Camagna and Andrea Forti, despite the low visibility, where able to record a short movie of some of the threes. We used this video for a fast 3D reconstruction, because it was particularly indicated, due to its characteristics: it was recorded for no SfM aims (as you see the movie sequences are not optimal for a 3D reconstruction), it represents the normal turbidity condition of the lake and it was done with an high lens distortion camera (GoPro Hero 2). For such reasons this material was perfect to hardly test SfM and IBM techniques for underwater archeology. In this animation you can see a short part of the movie (from the 15th second  to the 25th), which we used for the 3D documentation...

From "La foresta sommersa del Lago di Tovel" (T. Camagna, A. Forti)

... and here you can see the result:

I hope this post was useful. Soon, when the season will allow us to start diving again, we will go on with tests and experiment related with underwater archeology. I hope to write soon some new report about it.

SfM for Underground Documentation

Since last October Arc-Team is working on a challenging project about World War 1 on the southwestern frontline 1915-18.

Austro-Hungarian artillery piece in wintry position 1916

In a few years we will celebrate the 100st recurrence of the Italian declaration of war against the Austro-Hungarian Empire in May 1915.
For that reason the South-Tyrolean Heritage Departement was starting an extensive archaelogical survey campaign in high alpine environment up to 3900 m.s.l.
We've had to find out a rapid, lightweight and low cost method to document the uncountable amount of structures along both front lines.
The area of our pilot project is an approximately 6 sqkm large high plane between 2000 and 2300 m.s.l. called Plätzwiese. Between 1915 and 1917 it was a strongly fortified second line artillery position of the Austro-Hungarian Army.

Crew Cavern

Because of the seclusion of this area, today there we can find some of the best preserved residues of WW1 in Europe. We expect a total amount of more than 1000 archaeological remains just in this sector.
One of the major challenges is the documentation of the numerous tunnels and caverns driven in to the rock to protect troops and material from enemy artillery fire.

View towards the exit

We apply SfM to obtain pointclouds of the interiors, dealing with difficult conditions of illumination and space. After placing GCP's around the entrance(s) and measuring them with our DGPS, we start with one or more sequences of pictures in the entrance area, advancing along the tunnels, turning at their end and coming back to the starting point.

In the dark we use of course a tripod to allow long exposures and some torches, to illuminate shadow areas.
In this way we've taken up to 300 pictures for one single structure.
Various attempts of processing them in Python Photogrammetry Toolbox divided in a few single sequences, gave us some indication of the most advantageous way of photographing.

Point Cloud in MeshLab v1.3.2.

For good results we need to follow two important rules:

1. Redundancy: The same point has to be visible in much pictures possible.
2. Long baseline: To obtain correct geometrical information, the projection centers of the sensors have not to be to close together.

• Taking a just a single photo-sequence along one way of the cavern, causes often problems with redundancy and subsequently lead to empty areas in the pointcloud, especially along the ceiling of the cavern.

Recording strategy 01

• Another of the experiences gained, is that the software has big difficulties to match features form the photo-sequence taken on the way inwards, with features from the photo-sequence taken on the way back to the entrance of the tunnel.

Recording strategy 02

• We get the best results advancing on two parallel lines, as well on the way in, as on the way out of the structure.

Final result 01

Finally we processed the parallel sequences of each direction and referenced them after in Cloudcompare in the way how Ale Bezzi described it.

Final result 02

In June 2013, after the thaw, we will resume fieldwork in Plätzwiese and we hope to  acquire further experience.

Cloud distance tool.

I was working on different SfM/IBM of a grave we dug in 2010. we have the documentation of four different levels (see picture below). It was a complex archaeological context, with two skeletons buried in different times (double burial), both partially destroyed by the construction of the Renaissance apse. Moreover the tomb was built on the side of a prehistoric house.

I tried to rectify the point clouds inside CloudCompare v. 2.4 (normally i use GRASS with the ply importer addon or MehLab) and I discover this fantastic tool: compute cloud/cloud distance. It can calculate the distance between two different overlapping mesh, similarly to the GRASS command "r.mapcalc". As you can see in the pictures below, the distance analysis between the first and the last documentation can represent the quantity of removed ground. It could be really useful for analysis of damages in buildings.

first point cloud

fourth point cloud

cloud/cloud distance

cloud/cloud distance over the fourth point cloud

Extreme SfM: precarious situations and workplace safety

This post is a further contribution regarding the main benefit of SfM: its versatility. 

Thanks to its speed during the phase of data acquisition, this technique allows us to intervene quickly in the case where the archaeological excavation exposes precarious situations, often related with collapsing structures.

The image below is good example: we were working in the vicinity of a church, checking the excavator. The left pictures shows the part of a wall which was exposed during this operation, while the right photo reports the same subject just twenty minutes later.

The wall befor and after the collapse

Luckily, seen the precaurious conditions of the structure, we decided to document it with SfM and IBM techniques, and this is the 3D restitution of the wall before the collapse.

The 3D model of the wall

However, the documentation of archaeological arthefacts before their destruction, is just one of the benefits of SfM (mainly related with the speed during data acquisition). In fact the versatility of this technique is strongly connected with the hardware we need to collect the data: a simple digital camera. This make it possible to work without the necessity of a direct contact with object to be documented (especially if we can support the excavation with a direct reflex total station, to record the Ground Control Points needed in the post-processing georeferencing operations). This way to operate lead to minimize the risk and increase the safety in the workplace. The images below regard an example of this situation: sometimes it happen to be called to evaluate the damnages of costruction sites, already underway without the archaeological control. In these cases it can happen to document precarious situation, but, using SfM techniques, there is no need to stand in risky places (like under a section with many gravel layers), because all the operations can be done from a safety distance.

The gravel section

The 3D puincloud (thin points)

The 3D pointcloud (thik points)

 I hope this post was useful, have a nice day!