Project Details
Description
Project no.: 1299.
The aim of the present project is to describe three dimensional (3D) objects in order to model and simulate the shape variation. This allows for:
1. Knowledge driven design/optimal design.
2. Prediciton of the full object based on partial knowledge about the object (occlusion)
3. Prediction of a future object based on previous observations (growth prediction)
The methods in the study are based on morphometric tools also called shape analysis. The basis for the analysis is landmark data. Landmark are homolohues point presumed to correspond over the object of a data set. When having only a few landmarks the registration may be performed manually, but for thousands of points it becomes tedious and practically impossible. In many cases punctual landmarks are hard to establish in images, and the process requires considerable prior anatomical knowledge. We search for automated methods for landmark detection in this stydy. Such methods have already been developed in the Ph.D study by Per Andresen, but must be extended to provide a tool for industrial and medical 3D shape analysis. Having the landmarks defined for the complete data set, we are able to use well known methods from statistics such as Principal Component Analysis (CPA), different tests on the data (like test for distribution, effective dimension of the data etc.), model testing and validation etc.
The aim of the present project is to describe three dimensional (3D) objects in order to model and simulate the shape variation. This allows for:
1. Knowledge driven design/optimal design.
2. Prediciton of the full object based on partial knowledge about the object (occlusion)
3. Prediction of a future object based on previous observations (growth prediction)
The methods in the study are based on morphometric tools also called shape analysis. The basis for the analysis is landmark data. Landmark are homolohues point presumed to correspond over the object of a data set. When having only a few landmarks the registration may be performed manually, but for thousands of points it becomes tedious and practically impossible. In many cases punctual landmarks are hard to establish in images, and the process requires considerable prior anatomical knowledge. We search for automated methods for landmark detection in this stydy. Such methods have already been developed in the Ph.D study by Per Andresen, but must be extended to provide a tool for industrial and medical 3D shape analysis. Having the landmarks defined for the complete data set, we are able to use well known methods from statistics such as Principal Component Analysis (CPA), different tests on the data (like test for distribution, effective dimension of the data etc.), model testing and validation etc.
| Status | Finished |
|---|---|
| Effective start/end date | 01/08/1999 → 31/07/2001 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.