3D modeling as a tool for archaeological exploration

Using images or videos taken from manned aircraft and/or drones

Updated draft, May 6, 2018

NEW: explorations using images from drones in Ica, including Nazca, in progress. See Section VII.

I. Introduction

In the world of Archaeology there are several recent reports of discoveries that are being realized through the examination of remote imagery of different types. Here we only give a few examples of this:
(a) An ancient buried village was found using a drone in New Mexico, USA.
(b) "Improving archaeological prospection using localized UAVs assisted photogrammetry: An example from the Roman Gold District of the Eria River Valley (NW Spain)".
(c) Sarah Parcak: "Technology of the future is helping us save our past".
(d) Last is best: Lost Maya Megalopolis in Guatemala found using LIDAR technology (NatGeo, Feb.1,2018).

Aside from helping to discover archaeological sites, visualization and analysis of 3D models has for many decades been used in geological exploration and also aided in the discovery of spectacular natural landmarks. As a vintage example of the latter, the largest "natural bridge" in the world, was discovered in 1961 by using a stereoscope for 3D viewing aerial photographs taken by Peru's National Air Photography Service.

In America there are thousands of known ancient indigenous ruins in various states of preservation, and according to mainstream archaeology in Peru they cover various periods from around 5000 years ago (Caral), to the time of the Spanish Conquest between 1521 and 1535 or even later. Some of the best preserved archaeological sites are in the Cusco region, of which Machupicchu, located in the mountainous rainforest northwest of Cusco, is well known worldwide. The main reason for the good preservation of the famous Cusco sites is the superb craftsmanship of the buildings, whose masonry techniques remain a mystery as those of the Puma Punku site in Bolivia.

In Perú there are many legends about "lost cities of the Incas" in regions to the northwest (Vilcabamba) and east of Cusco (Paititi) which have not been unequivocally found to date in part due to the dense forest that has covered many possible structures dispersed in thousands of square kilometers that have not been sufficiently explored. In addition we have increasing evidence that there is a hidden population living in the plateaus of the northern Vilcabamba mountains, perhaps the descendants of the Incas themselves, which could be the peruvian counterpart of an until recently unknown indigenous civilization found only a few decades ago in the rainforest of the Sierra Nevada de Santa Marta. A very interesting conference was given in June 2014 at the Google Outreach offices in California with two of the leaders of one of the four ethnical groups living in the Sierra Nevada. This almost hour long conference includes presentations by different experts of which the brilliant geographical description by the chief cartographer of the Amazon Conservation Team (ACT) is especially noteworthy (minutes 8:52 to 22:55).

With modern technological tools it is now possible by air to effectively discover hidden structures (Images 1 and 2), and measure them with sufficient accuracy using 3D Photogrammetry, for determining whether these structures are natural or man-made, all of this before venturing into the much more expensive verification on land. In other words, investment in exploration can easily be justified due to the low cost of the new methods and the speed of its realization.

The remainder of this essay partially recounts an aerial exploration experience of 2012 in the middle of a dense mountainous jungle in Peru, includes an example of 3D reconstruction and measurements using Pix4Dmapper Pro with images obtained from a video clip in that flight, analyzes a few of the anomalies detected therein, makes some recommendations for future exploratory strategies, and in the Appendix presents a few interesting anaglyph images which require special glasses for visualization.

II. Obtaining the Data

In July 2012, a Cessna U206G single-engine plane piloted by Enrique Tantte took off from a small airstrip in Satipo for a scouting mission looking for hidden archaeological remains in the highlands of the Cutivireni river basin. Antonio Gómez, Jorge Gómez (†), Jorge Mattos and a representative of the Otishi National Park headquarters in Satipo contributed in this expedition with photos and videos. The Cutivireni river is the main tributary of the Ene river. The objective was to take still images and videos of an area of around 400 square kilometers on plateaus with altitudes between 2200 and 2600 meters above sea level that had been partially explored in previous aerial missions, having found a lot of traces of ancient human activity in the region but nothing conclusive relating to an important "lost city". Image 3 shows a Google Earth map with the flight path registered by a Garmin GPS, and a yellow rectangle showing the planned area of interest. At that time we did not have any knowledge of the new 3D reconstruction methods so the flight was not optimized for this type of image processing.
Two spectacular anomalies were accidentally discovered in this flight. Image 4 and Image 5 by photographer Antonio Gómez show these features. You can click on each image for a larger view.

Unfortunately insufficient images were taken of these features in the number and manner that would make it possible to realize a 3D photogrammetric reconstruction from them. However some pairs of images enabled us to create stereo-anaglyph images of the objects in images 4 and 5.  In the Appendix of this essay we show these anaglyphs and those of other interesting features in the region.
Fortunately, upon examination of the flight path and the videos of the expedition, we found an interesting feature within a small loop in the southernmost part of the flight where a Full HD video was taken by Jorge Gómez in a favourable trajectory and angle for obtaining a 3D model. Image 6 is a Google Earth image of the area selected for the 3D model, part of the GPS track and the surrounding general area showing a few of the hundreds of traces of possible ancestral intervention by human beings of the Megalithic Age in the region. Extracting 209 frames from a 70 second clip from that video, enabled us to perform an adequate 3D reconstruction of the feature, that consists of two very steep and narrow connected gorges in a plateau mounted over over a larger plateau, acting as a water collector, at a distance of only one kilometer from the cavern shown in Image 5. The video was taken through a flight path forming an almost circular arc with a radius of about 600 meters, at a height over the terrain of around 400 meters. Image 7 is a Google Earth image zooming in on the area of interest and showing the camera positions of 36 of the 209 extracted images.

III. Processing and Analysis

-----IIIa: Using a useful new tool in Pix4Dmapper version 2.0, 209 equi-spaced frames were selected from the mentioned Full HD video clip for constructing the desired 3D model. Without this tool this job can be done manually but in a very laborious manner using special video processing software. The "Image Properties Editor" (Image 8) shows that the 209 images were loaded successfully. The camera was not identified. There were 11 blurred images in the set, which were disabled before processing.

-----IIIb: "1. Initial Processing" and "2. Point Cloud (without Mesh)" are run with default settings before georeferencing. Image 9 shows the dense point cloud. Using the "Area Processing" in the next run will eliminate extraneous points in the Cloud.

-----IIIc: Georeferencing and Scaling is performed using Google Earth as reference. This is a painstaking job which demands much operator experience in order to minimize errors. These are due to the difficulty in identifying common points between the Google Earth image and the Point Cloud in part due to the low resolution of the images and also due to the complexity of the jungle landscape. Image 10 shows the initial (sparse) Point Cloud after georeferencing with 3 ground control points (GCP), 4 manual control points and 1 scale constraint to the original Point Cloud and limiting the Area Processed to a polygon whose area is approximately 109 Hectares. The Quality Report shows some parameters of this result.  As can be seen, the average error in GCP is around 2.7 meters which permits us to do sufficiently accurate measurements for the purposes of this report. A short video clip shows that the KML orthomosaic fits satisfactorily on the Google Earth map confirming the fact that the georeferencing is adequate.

-----IIId: The result of running the process of densification and texturing ("2. Point Cloud and Mesh") is shown in Image 11.

-----IIIe: Preliminary visualization. The following 3D animation gives a general idea of the topography and the colors. Notice especially the size of the cliffs in comparison with the Google Earth version and the difference in colors of the Secondary Forest (bright green) compared to the Primary Forest (darker and grayer green).

-----IIIf: Measuring features

IV. Conclusions

-----A. Though there were problems of visualization due to the low resolution of the images, dark shadows and/or lack of sufficient angles for camera triangulation, the quality of the reconstruction was sufficient for an adequate characterization of the main anomalies in the area used in the 3D reconstruction. We did 3D and also 2D visualization since both systems offer peculiar and complementary advantages. Visualization by trained human eyes makes it possible to detect a series of details in complex images which computer processing incapable of achieving for the time being, no matter how sophisticated the algorithms used. Obviously this capability is tremendously enhanced with the modern computational tools that are partially described in this document.

-----B. Though the features selected for showing the method did not constitute a definite archaeological discovery, we did find anomalies suggesting probable human intervention, whose age is of course unknowable at this stage. The exercise demonstrated the power of the methods used, in spite of the low resolution (only 2 megapixels) of the video frames used in the 3D reconstruction. Today consumer grade video cameras have 4K capabilities which is equivalent to 8 megapixels, that is four times better resolution than the images used in our example. With this enhanced resolution and an adequate flight plan, it should be possible to obtain a much better 3D reconstruction than the one described in the previous section, and thus arrive at more reliable conclusions regarding the nature of the features observed.

-----C. The resulting 3D model is vastly superior to that obtainable in Google Earth (GE), in spite of the fact that the ground sampling distance (GSD or pixel resolution = 45cm) in the horizontal sense is of the same order of magnitude as in GE. The basis of the superiority lies in the much better vertical resolution, where GE fails completely for small objects since it uses only one satellite image for their visualization, which is draped over an independent digital elevation model (DEM) obtained from the Shuttle Radar Topography Mission (SRTM) of the year 2000. Since November 2014 this DEM for Peru has a GSD of 30 meters for some areas, but in spite of an important improvement from the previous 90 meter pixel, it is still impossible to measure vertical distances in GE for small cliffs or vertical objects. On the other hand, the point cloud method based on multiple images makes it possible to obtain vertical resolutions that are of the same order of magnitude as the horizontal resolutions, thus enabling the viewer to see the sides of objects from different angles. This is not possible in GE, except possibly (though distortedly) for very large features whose sides (for example at angles over 45 degrees) are very large, say 50 meters or more. An example of this possibility is given by an extraordinary feature we discovered a couple of years ago at less than 15 kilometers south of the feature that we used in our example. This discovery has not been confirmed, since to our present knowledge it has not been photographed from a close enough range.

-----D. The analysis in videoG suggests possible human intervention in the small area containing two specific features, a 670 square meter platform and 890+ cubic meter mound right next to it. In addition, across the narrowest part of the southern gully, a bright green vegetation showing Secondary Forest can be seen, including a straight object more than 6 meters long that overhangs the cliff (See Image 13). In the plateaus and deep gorges surrounding this area there is an uncountable number of places with secondary forest, of which the most notable one is the one higher up on the plateau at about 1900 meters to the south east of the reconstructed area (Image 14).

-----E. As can be seen in section IIIf1, the oblique images do not have to complete a circle (or ellipse) around the features to be reconstructed using Pix4Dmapper, in order to obtain results that are adequate for discovery, even for approximate quantitative measurements. A further example of this is given in another reconstruction using 52 frames from a video taken in 2010 from a helicopter. This may be especially useful in situations where the features of interest are located in the sides of high mountains, for example inside of a steep valley, since it would be impossible for the aircraft to circle the mountain at an appropriate altitude for obtaining the desired resolution. An interesting example of this could be the famous Ollantaytambo site in the "sacred valley of the Incas" of Cuzco. Here the main archaeological constructions are at altitudes above sea level between 2860 and 3000 meters, while the top of the mountain is over 4000 meters. A reasonable flight plan here could be a couple of arcs or semicircles the largest of which would be at around 3200 meters and the smallest higher up at around 3300 meters, with the necessary precaution of keeping a safe distance from the cliffs.

-----F. Manned AirCrafts (MAC) vs Drones (UAV): which are applicable for our objectives? In the case of the mountainous amazonic region of our study, clearly MAC is the only option for now. The reason is that the region of archaeological potential - the plateaus and the higher parts of the Cutivireni River Basin - is far away from any possible access by land so the only immediate options are overflights by airplane or helicopter. Image 15 illustrates this point. The red highlighted areas shown in this image, spanning approximately 1000 square kilometers, include a multiplicity of possible target sites for exploration such as those shown in the first two chapters and in the Appendix of this study. Access by land will be feasible when areas apt for helicopter landing (or rappel) are identified after the first overflights are accomplished with the corresponding detailed photogrammetric surveying using methods as described above.
-----Once the more interesting archaeological areas that can support helicopter landing are identified, and the permissions are granted by the authorities, drones will be the best option for obtaining the best quality of 3D models, orthomosaics and topographical models for a detailed characterization of the respective sites.
-----If access can be achieved into at least some of the caverns, they can be efficiently mapped using modern surveying systems such as Field Map. As modern photogrammetric systems based on multiple images keep improving, spectacular 3D models can be obtained of these caverns, perhaps using infrared cameras with that can "see" in the dark or even spherical (360°) cameras.
--

V. Planning future projects

-----Image 15 shows that there are only two airports, Mazamari and Malvinas, close enough to the target areas with archaeological potential mentioned in this study, from which aerial expeditions could be realized advantageously. Mazamari is approximately at 140 Km and Malvinas at 76 Km in a straight line from the center of the mentioned areas. Using a single motor Cessna airplane would be most convenient, since the wing of these airplanes is over the fuselage and does not interfere with the video and image capture. Logistic and economic factors should be analized and most importantly, the weather should be closely monitored, due to abundant seasonal tropical storms and even sporadic rainfalls in the "dry season".
-----Though the Mazamari airport is around 40 minutes away from the area of the exploration targets, whose locations are well identified from our previous aerial expeditions, a total of 140 minutes should be sufficient for each overflight from Mazamari (one hour for circling a couple of targets) if the weather is OK and the flight plan and image capture is adequately executed. This airport has the advantage of being at 20 minutes distance by land from the headquarters of the Otishi National Park, that is the authority in charge of the protected areas that comprise the targets of our proposed explorations, one of whose representatives must be a part of each expedition over the Park. Also, though the Malvinas airport is much closer, this airport is dedicated to servicing the administration of the natural gas installations of Plus Petrol in Camisea and Malvinas, so it would be more difficult to adequately plan the flights from there.
-----The Kiteni airstrip is a third possibility, but though it is closest to two of the most atractive features in the southern flank of the Cutivireni plateaus, reaching Kiteni entails the greatest cost. The main advantage of this location is that modern helicopters operate there, used in the monitoring and maintenance of the gas pipeline that goes from Malvinas to Pisco (a second pipeline transports most of the gas to Lima). Using a helicopter instead of an airplane offers definite advantages for image capturing, but is much more expensive.
-----A critical aspect to be planned is the required close coordination between the Company or Association that designs the expeditions and the pilots of the airplanes or helicopters in which they will be carried out. The flight over the predefined targets must be carried out as closely as possible to the flight plans, otherwise the quality of 3D reconstruction will be hampered. The flight plans must be discussed at length with the pilots in order to take into account technical constraints of the flights.

--
VI. Epilogue for the study of the Cutivireni plateaus in the mountainous jungle in Perú

----- The main concern which motivated this study is the proliferation of narcotraffic and terrorism in the Ene river, part of the VRAEM system of valleys, at around 60 kilometers from the areas described above. This unfortunate situation is probably the main reason why the eco-touristic potential of the region has not been developed until now, this being its only sustainable alternative for development given its extraordinary natural landscapes and archaeological potential. If we could organize one flight with a light airplane and make a couple of important discoveries with the methods described in the above report, we would attract international attention and financing for further expeditions and archaeological studies.
-----Once the archaeological potential has been confirmed, and the approval of the neighboring native Ashaninka and Machiguenga communities of the area has been secured, then interdisciplinary studies of eco-touristic projects can be carried out for fostering the sustainable development of the region, with adequate environmental and social management. This could eventually help to turn the tide against the widespread corruption that is ruining the moral fibre of the Nation. We need help from people that are sufficiently intelligent and conscious of the need to protect the environment and thus the native people, but this cannot be done without well planned projects that can be adequately financed. This is a long term view, but my concern is that the more we delay the more damage is done to the environment, and we would lose another battle against Corruption, with all the dire consequences that this entails.

VII.
VII. Explorations using Drones in the Southern Coastal areas of Peru, mainly including the Ica Desert

-----Notwithstanding the great potential importance of the area of the Cutivireni river basin covered in the previous sections of the present report, there is another region in Peru that has perhaps a greater potential for tourism. Its main features are centered around a very famous and mysterious archaeological wonder, known worldwide as the Nasca and Palpa lines and geoglyphs located in the south-eastern part of the Ica Region in Peru (see political map). A great advantage of the Nasca (*) and Palpa zone is that it is readily accessible due to its basic existing infrastructure (see general tourism Map in a preliminary draft from Google Earth). (*) Note that in English speaking countries Nasca is still referred to as Nazca, but this has changed in Perú, and also in the UNESCO reference.
-----In this section we will show a few results obtained with images and videos taken with quadricopter-type drones, an effective method for developing photogrammetric 3D models that are very useful for exploration purposes and detailed analyses.
-----As a general background, the great desert in the Ica Region of Peru, covering an area of more than 10,000 square kilometers, contains some of the most important archaeological and paleontological zones of Peru, and even of the world. Searching with Google under "Touristic potential of the Ica Region in Peru" we find 585,000 results, of which the following links will start giving us an idea of the yet untapped but extraordinary potential of this region:
-----(a) Peru: Nazca's to-be-unveiled tourism potential | Noticias | Agencia ...
      (b) Ocucaje: Buried in Peru's Desert, Fossils Draw Smugglers.
      (c) Paracas is an already developed tourist center that is at around 210 Km by highway from the city of Nasca.

Model VII-1
      The area of this model is approximately 11 hectares centered around the following coordinates:
14°10'29.32"S; 75°50'58.15"W. This location is at 56 km southeast of the Paracas tourist center and 90 km northwest of the Nasca lines zone, as the crow flies. The access is from the city of Ica in two sections: the first is along a paved highway westward for 14.8 km and the second from location (14°6'31.22"S, 75°51'0.56"O) through a dirt road for approximately 8.4 km.
      For this model we show two images.
      Image vii-1a is divided into two parts. Part A is a view from Google Earth of the area, and part B is an oblique 3D view from the Pix4Dmapper software showing the relative positions of the photos taken with the drone using the Drone Deploy App for automatic flight and image retrieval.
      The following image shows a surprising finding that is an example of the potential of modern photogrammetric technologies for new discoveries in geological and archaeological exploration.
      Image vii-1b shows the Orthomosaic and the Digital Surface Model (DSM) obtained with Pix4Dmapper Pro and exported to Global Mapper. The DSM shows a large nearly circular depression whose diameter is estimated to be over 500m since the area covered by the model is clearly not large enough. If this depression was not caused by a large meteorite, it could be an indication of an artificial underground site which should merit further investigations, perhaps including careful excavations or other modern underground detection systems. The point here is that the great Ica desert holds an immeasurable number of important buried archaeological and paleontological sites, and fundamentally requires a thorough revision of the geology of the region, including accurate dating of the various strata with double checks using modern equipment and techniques.

Model(s) VII-2 (this case study is part of a critical regional problem which needs to be studied in depth)
      The following models of one of the flat hills in the Palpa Province of the Ica Region were obtained using images taken from a DJI Mavic Pro drone on July 30, 2017. The basic model, using 18 vertical (nadir) 12 megapixel images, covers 34 hectares which is enough to include the flat area of around 7.5 hectares and surrounding slopes, and has a GSD (ground sample distance) of 8.1 cm/pixel. The larger model uses aditionally 120 frames (8 megapixel) from two 4k videos, covers 107 hectares and its GDS=11.9 cm/pixel. The smaller model is "cleaner" (less noise) and is the one used for the analyses done in the linked essay, which is an ongoing attempt to understand the real significance of this hill (or mountain if you please) as a model for explaining many of the mysteries of the Nasca and Palpa geoglyphs.
      Next we show here the still views of both models together with the positions of the images used, and present the respective video animations. When possible we plan go back to the site and shoot closer images of the more interesting areas in order to construct 3d models of much greater detail, perhaps under 3cm/pixel. This was not done in our previous visit since our initial objective was to get a complete view of the hill with a reasonable amount of photos.
      Still images: SiBasic Model and SiExpanded Model.
      Video animations: VaBasic Model and VaExpanded Model.

VII-3
Models VII-3: Photogrammetry in Nazca.
      In February and March 2018 our friends of TheDrone Peru have made several drone flights in the zone of the world famous geoglyphs. One objective was to examine in detail the sensational discovery, published in Dec. 10, 2017 by José Luis Camacho of the "Mundo Desconocido" YouTube Channel, of 18 "circles" that are mysteriously aligned for 18 kilometers and extrapolate to what seems an ancient cemetery. We have modeled the last two (17 and 18) that are sufficiently close to the highway (within the range of modern DJI drones), using rigorous photogrammetric methods. The result can be seen in two images obtained by processing pre-planned sets of nadir images with Pix4Dmapper: Circle_17 and Circle_18. Both show almost exactly the same diameter: 8.33 meters for the outer ring. We have published a YouTube video (9:08 min, in Spanish), on these circles showing other contextual aspects not previously covered, but this is an unfinished job whose purpose is to highlight the need to protect the area from further depredation, and promote sustainable development of the Region. A second video on these circles (6:40 minutes, in English) shows new measurements and a discovery of surprising internal properties of the circles made by a Carlos Escobar, a geologist who is collaborating with our investigations.
      A second project that we have published based on a flight with a drone, is a photogrammetric analysis of a famous geoglyph of a Tree located right next to the PanAmerican highway. In this short video (4:44 minutes) we show the steps done to obtain various measurements of the geoglyphs, the last two of which cannot be done with the present satellite images due to their insufficient resolution. A similar job can be done for all the rest of the lines and geoglyphs of Nasca and Palpa, but instead of using a drone, we recommend taking the images from an airplane for a much faster and non-invasive capturing of the hundreds of thousands of images required. This is a job that requires much planning and coordination with the authorities, but which we consider essential for the preservation and restoration of the damaged geoglyphs. The extraordinary power and low cost of the modern surveying technologies gives peruvians a good opportunity to do this job in a fairly short time. Speed is essential in this work given the critical situation, not only of Nazca but of the country as a whole.
 

Models VII-4: Videos of other 3D models from southern coastal areas in Perú, to be analized in future updates.

Ica dunes, Huaca El Salitre, La Centinela ruins complete, La Centinela ruins partial
---------------------------------------------------------------------------------------------------------------------------------------------    

APPENDIX. Examples of the use of stereo anaglyph images for analyzing some features of interest in the Cutivireni plateaus. (Note: the use of Red-Cyan glasses is recommended, instead of Red-Blue.)

-----A1. Possible hidden structures of Images 1 and 2 (Image A1).

-----A2. Tower of Image 4: The two photographs used for this anaglyph image were taken too close from each other so the depth of the tower and especially that of the broken piece in the front are not optimally represented in this anaglyph (Image A2).

-----A3. Cavern of Image 5 (Image A3).

-----A4. Great anomaly (Image A4) at less than 12 km distance from the features reconstructed in Chapter III of this report. The anaglyph was made from two vertical (nadir) photographs taken by the National Air Photograph Service of Peru. There is a vertical exaggeration in the view since the pair of photos for the reconstruction were taken far apart from each other.
--

BIBLIOGRAPHY

P. A. Rizo-Patrón: The Ene River Basin in Perú - its geography and potential for tourism,
in "The nature and culture of Latin America : review of Polish studies", Zbigniew Mirek, Adam Flakus, Andrzej Krzanowski, Andrzej Paulo & Janusz Wojtusiak (Editors), W. Szafer Institute of Botany, Polish Academy of Sciences, Kraków, 2010, pp. 105-134.
For a PDF copy of the published article Click here (includes hardcover image, book and copyright information, Foreword by the Editor, Introduction in Spanish, Contents of the book, and the complete article - 15.9 megabytes). For printing or exporting to other users please contact the author.

Email us