Fossils Unearthed Using Artificial Neural Network Software
11/22/2011 by: Darleen Hartley
Computer networks that imitate how the human brain works have been pinpointing fossil locations. An artificial neural network (ANN) developed by a research team was tested in the Great Divide Basin of Wyoming, a 4,000 square mile of rocky desert. The goal: to make anthropologists' scrutiny of large areas less time consuming and more productive.
Dr. Glenn Conroy, paleoanthropologist, reknown for discovering the first known Miocene hominoid (Otavipithecus namibiensis) ever found in southern Africa, was joined by Western Michigan University PhD colleagues Robert Anemone, a biological anthropologist, and geography professor Charles Emerson who has an extensive knowledge of satellite imagery.
The software takes into account infrared electromagnetic radiation, satellite images, and maps to determine elevation, slope, terrain, and other landscape features. With detailed input of existing fossil containing areas, the ANN is used to locate other potential sites. Pattern recognition looks for similar land features. Their discoveries are "a combination of hard work, meticulous planning and a good dose of luck" according to Conroy.
Six bands of visible and infrared light recorded by the Landsat 7 satellite were the basis of the study. Anemone assigned pixels to forest, scrubland, and wetland. He marked the pixels to identify a fossil or non-fossil site. Most fossils were located in sandstone. The network was thus trained to identify fossil yielding areas.
It takes a trained eye to separate rock from fossil at sites tagged by the ANN
In the test area, the ANN identified 79 percent of the already known sites and correctly tagged 99 percent that contained fossils in the area that has already given up early mammal fossils from 50-70 million years ago. Moving farther afield to the Bison Basin of Wyoming, the ANN was able to identify four fossils sites in similar, but unfamiliar, terrain.
The ANN will be going overseas to look for early hominid fossil sites on the continent where Conroy's amazing discovery occurred. Conroy has conducted explorations across the world, including Kenya, Argentina, and Pakistan. He applies computer imaging techniques to explore anatomical structures in extinct primates and fossil humans. Conroy has written numerous articles with such intriguing titles as: Endocranial features of Australopithecus africanus revealed by 2 and 3-D computed tomography. In the US, Dr. Anemone's research in the Great Divide Basin focuses on the effects of climate change on the evolution of primates and other mammals.
Anemone said: "In the time we are working with - 50 million years ago - there was a major event of global warming. The earth's climate was warmer than what it had ever been, so we are interested in the effects of climate change in the past on living things so we better prepare for climate change today by seeing the past events."
It takes human sweat, in combination with AI to locate and tease fossil bits from the earth
One qualifying factor built into the ANN software was that the land had to be sloped by at least 5 degrees, so that erosion would have exposed portions of the fossil bed. July 2012 is planned for an on-site excursion into the Wyoming areas tagged by the ANN as potentially hiding fossils. "We're going to go to some areas we've never been to, that we wouldn't have been aware of, and see what we find," Anemone says. Boots on the ground are necessary to identify bones from among rocks in an outcrop. Although they are using technology, specifically AI, to locate the sites, Anemone says this eyeballing "is something robots are not yet ready for."
robot, ANN, Wyoming, Great Divide Basin, 3D, computed tomography, Bison Basin, AI, fossils, Australopithecus africanus, Glenn Conroy, paleoanthropologist , Miocene, hominoid, Otavipithecus namibiens, Western Michigan University, Robert Anemone, anthropology, Charles Emerson, satellite image, Artificial Neural Network, Landsat 7
© 2009 - 2011 Bright Side Of News*, All rights reserved.