An Investigation Into Techniques for Landmark Identification of 3D images of Human Subjects. Phase 1
Abstract
Anthropometric data have been used by the Air Force for many years to help in the design of clothing, equipment, cockpits, etc. The use of anthropometric statistics enables designers to provide a better-fitting product and ensure that the people who will be using the equipment are physically able to do so. The Human Engineering Division has been exploring the use of 3-dimensional scanning technology to produce a digital representation of the surface of the human head. A vertical stripe of laser light is projected onto a stationary subject as the scanner rotates 360 degrees around him or her, taking video recordings which are converted into digital format. Although the data collected in this way are much more complete than was possible using traditional methods, techniques have not yet been developed to easily analyze and use this type of data. Several approaches to landmark identification using artificial intelligence methodologies were investigated. A blackboard architecture was identified as the most promising approach. Several prototype modules were developed to generate and evaluate hypotheses regarding the positions of the landmarks on 3-D digital images. These were tested on a sample of 20 subjects and the results were sufficiently encouraging to warrant further work. The design of a blackboard system was begun, including components utilizing other technologies such as neural networks and constraint networks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1989
- Accession Number
- ADA218614
Entities
People
- Randy B. Pollock
Organizations
- Universal Energy Systems