Image of a humanoid geoglyph (left) and an enhanced image (right). (Image Credit: Yamagata University)
A group of scientists has uncovered 143 unknown “geoglyphs” in Peru by using artificial intelligence and satellite imagery. The ancient drawings, discovered in the Nazca desert, dated between 100BC and 500AD and were created by removing the rocky black topsoil of dark stones to reveal light-colored sand underneath that depict different geometric shapes, people and animals.
A team led by Masato Sakai at Yamagata University in Japan began to map out every line of geoglyph to preserve and understand their significance. Through the analysis of high-resolution 3D images and over a decade of onsite work, the team discovered 143 new drawings; they announced their findings on November 15. The geoglyphs are split into two main categories, Type A and Type B. Larger animal-like drawings, as long as 100 metres, showing sites of ritual ceremonies are categorized as Type A. Type B contain smaller drawings, as small as 5 meters, that were usually found on slopes and paths, possibly serving as a waypoint for people who were traveling.
A processed image of a two-headed snake geoglyph. (Image Credit: Yamagata University)
There are more geoglyphs that have not been found yet, but it has been more difficult to find them. Most of the lines in the drawings have sustained damages from floods, roads, and urban expansion. The team believes the drawings are also located in other areas, which would take additional time to discover using traditional methods.
“The expansion of urban areas in Nazca has brought damage to the lines, resulting in a trend that has drawn attention as a social issue,” according to the statement. “There is an urgent need to gain an accurate understanding of the geoglyphs' distribution so that work can be done to protect them.”
In a collaborative effort with IBM Research, the team is now using deep machine learning algorithms to help with discoveries. The team is also using a cloud platform to put a large amount of geospatial data together, which includes LIDAR, drone, and satellite imagery and geographical surveys. This is to create highly detailed maps of the search areas.
The researchers then trained a neural network to recognize data patterns of known lines to search for new lines. While testing the neural network, they discovered a new drawing, a small humanoid, which had not been found in previous data collections. By using this method, the team was able to make the discovery in just two months, whereas other methods had usually taken years.
Have a story tip? Message me at: cabe(at)element14(dot)com