AI identifies 3 new Nasca lines hidden in Peruvian desert

Machine learning has uncovered previously hidden 'Nasca lines ' geoglyphs etched into Peruvian desert 2,400 years ago.
John Loeffler
A Nasca geoglyph depicting a pair of legs
A Nasca geoglyph depicting a pair of legs

The Yamagata University Institute of Nasca 

The Nasca lines of Peru are one of the most mysterious ancient archeological sites in the world, but thanks to a new AI technique, researchers may have a new way of identifying hidden Nasca lines that have so far evaded detection, including three new Nasca lines that might otherwise have been missed.

Nasca lines are ancient geoglyphs etched into the Nasca desert in Peru (as well as a few other sites) and are believed to be between 1,400 and 2,400 years old. Their purpose is a mystery, and it is believed that two different cultures: the Nasca culture, which is dated to about 200 BCE to 600 CE, and the older Paracas culture, though a majority of the known Nasca lines are a product of the former. The lines themselves are thought to have been made by using black stones incising the surface of the desert to reveal the white sand underneath, according to LiveScience.

Being more than a thousand years old, erosion has obscured many of the geoglyphs, making high-altitude aerial surveying a challenging process. That's where Masato Sakai, a professor of anthropology and archaeology at Yamagata University in Japan (YU), comes in. Partnering with IBM Japan, Sakai and his colleagues at YU Institute of Nasca have turned to AI to help analyze high-resolution aerial photography to identify key features that could be geoglyphs in disguise, rather than natural landscape features.

"The known geoglyph patterns are unique and complex," Sakai said in a YU statement on the new discovery. "Therefore, it is likely that new geoglyphs will not have the same design as the existing ones. It is difficult to find new geoglyphs using deep learning object detection models trained only on known geoglyphs, as they may not be able to find features that do not exist in the training data."

"To address this issue, we divided the known geoglyphs into relatively simple pictorial elements and used them as training data to create an object detection model," Sakai continued. "We hypothesized that there would be similar elements in new geoglyphs and the method may improve the generalization performance (the ability to detect new geoglyphs on previously unseen data) by focusing on capturing these elements rather than the entirety of geoglyphs."

The method seems to be working, helping Sakai and colleagues identify four previously undiscovered Nasca lines depicting a humanoid, a pair of legs, a fish, and a bird, which were then verified by an on-site inspection by archaeologists. The findings were detailed in the most recent issue of the Journal of Archaeological Science.

The hunt for Nasca geoglyphs is a pressing one, since erosion, climate change, and human activity are accelerating their erasure from the desert's surface. If they are to be documented and preserved, they have to first be found. According to Sakai and his team, the new technique was 21 times faster at identifying new geoglyph candidates than a trained archaeologist using the naked eye alone, and should dramatically speed up the process of documenting these ancient wonders before they're gone for good.

Add Interesting Engineering to your Google News feed.
Add Interesting Engineering to your Google News feed.
message circleSHOW COMMENT (1)chevron
Job Board