Wednesday, October 16, 2013

Drone Mapping the Pyramids of Zuleta


Drone derived slope model of portion of Zuleta.
The Pyramids of Zuleta are one of the hidden treasures of the Andes.  Built around 1,000 years ago, by the native Caranqui people, these earthen mounds and platform pyramids dominate the landscape near Hacienda Zuleta.  Unlike much of our planet, high resolution aerial imagery and digital elevation models are unavailable for this part of the world.  This is due to the fog that often blankets the area and the agrarian nature of the region. As a part of a team of archaeologists who visited the site in the August 2013, we aimed to change that.

Comparison of Landsat Imagery (best public data available) vs. UAV
Using a small, hand-launched, Unmanned Aerial Vehicle (UAV aka drone) equipped with a downward facing camera and a sophisticated autopilot system, we documented the site as it has never been seen—from extremely low altitude and at high resolution.  This was a challenging task as most of the pyramids are located in the bottom of steep constricted canyon inhabited by Andean Condors.  To make things more challenging there were high winds, clouds, and quirks of the micro-climates within the canyon to contend with. In spite of that, we were able to fly nine missions and collect hundreds of photographs in just a couple of days.
Programming the UAV while in flight.
The UAV flies in a defined pattern and as it collects photographs. The onboard autopilot insures that each image has 60% or more overlap with adjacent images.  These overlapping images allow for the data to processed into 3D and digital terrain models (DTMs) using photogrammetric and cutting edge Structure from Motion technologies.  All of the data are GIS ready. While basic processing allowed us to see the mapped data in the field, it was necessary to further develop it using a high-end processing farm in Maryland, once we were back in the States.

Radiance scaled enhancement of UAV data.
This approach has already led to the discovery of many more mounds that are not obvious to the naked eye but stand out in the data.  Furthermore, it is the first time the detailed spatial relationship between each of the earthen structures can be explored with precision.  This data will be used to track the condition of the mounds over time and has created a digital snapshot of their current state for future generations to ponder. This was accomplished with only a few days of fieldwork and under harsh conditions.  To create a map of similar accuracy using traditional survey approaches would have taken weeks and lacked the aerial imagery this approach provides. 
Pyramids deep within the canyon.
Our team has conducted similar mapping missions in Australia, Belize, Belgium, Fiji, France, Germany, The Republic of Kiribati, Peru, and South Africa. We look forward to continuing these sorts of projects elsewhere and helping to preserve and understand our past.


Hand launching the drone.
The archaeological mapping and processing team included Chet Walker and Mark Willis.  The rest of the archaeological team was made up of Steve Athens and David O. Brown. Athens and Brown are experts on the archaeology of the Andes and have worked for decades in the region.  We would like to thank Fernando Polanco Plaza, the general manager of Hacienda Zuelta, for his hospitality and dedication to preserving the rich heritage of Ecuador.



10 comments:

  1. Have you tried processing with Pix4uav software from pix4d.com ?

    ReplyDelete
    Replies
    1. Hi Roberto, I have not processed this particular dataset with Pix4UAV but I've used the process they based that software on to make the 3D models. I have tested Pix4UAV in the past and it is robust.

      Delete
  2. Mark,

    Congrats on the coverage in Gizmodo, awesome story and very interesting work. I cover Texas for Amazon Web Services and we have a variety of companies that use us for massive image storage at very low costs. If you are interested in looking at this in greater detail, send me a message and I would be happy to chat.

    ReplyDelete
  3. Looks great. What UAV/autopilot platform did you use?

    ReplyDelete
  4. Fantastic Mark and Chet!! with this work we are step closer to our goal, uncovering the mysteries of the Karanki world!! Thank you to you all!! I send you VERY good energy from the Andes!

    Come back!!

    Fernando

    ReplyDelete
  5. Wow grate stuff, what UAV/RPA are you suing, I have Sensefly's Swinglet and Ebee for survey mapping and they are fantastic, but yours looks a little different, what is it?

    ReplyDelete
  6. Have you tried to process the imagery with APS by Menci Software?

    ReplyDelete
  7. Cordial Saludo

    para consultarte hacerca de un proceso que estoy llevando a cabo con lps y fotografias de baja altura (150 m). los puntos de control han sido tomados en campo y las fotografias continen informacion considerable de vegetacion. sin embergo, los ajustes de en la aerotriangulacion no son los que esperamos obetener.

    gracias

    ReplyDelete
  8. Thanks for this blog. Just saved me a heap of money on software. Trying to make point cloud from pole shots on an archaeological survey in Australia. I knew it could be done. but I didn't know how cheaply! Thanks so much for sharing your knowledge.

    ReplyDelete
  9. following your Electronics-UAV realted sites I am very glad follow me on:-If you are really Aerial Videography drone interested and find enough details in UVA the particular article will help you to dji autopilot systems determine the impact and the effect it has spread in the field and by which you can carry on with all day to day works and the amazing Drones for sale changes that science is bringing in our life with the new systems like Aerial Photography drone.

    ReplyDelete